Peter Yang

Ok I've fallen for the ragebaiting and have been triggered by this so I'm going to say this: Whether intentional or not, there's a lot of virality that comes from: 1. Pick the most ragebait moment of the conversation 2. Make a clip of that exact moment 3. Get tons of retweets and views I can tell it's rage baiting (just like the Corgi founder clip) but here's the thing - if even one parent watches this clip and decides to neglect their kids more then the damage is done. And yes it's hard to work at home when your kid comes back at 3 pm (or 1 pm with summer camp!) but we all make our choices.

好吧，我还是被这种 ragebait（钓怒、故意激怒人博流量）的内容钩到了，也确实被触发了，所以我想说这个：不管是不是故意的，很多传播性其实都来自这样一套做法：1. 挑出对话里最容易让人愤怒的那个瞬间 2. 把那个瞬间单独剪成 clip 3. 收获大量 retweet 和浏览量。我看得出来这就是在做 rage baiting（就像那个 Corgi founder 的 clip 一样），但问题在于——哪怕只有一个家长看了这段 clip，然后决定更少照顾自己的孩子，伤害就已经造成了。没错，当你孩子下午 3 点回家时，在家工作确实很难（如果参加 summer camp，甚至下午 1 点就回来了！），但我们都是自己做选择的。

I'm reading this and I still don't get what a dynamic workflow is or when to use it in Claude Code

我在读这个，但我还是没弄明白 dynamic workflow（动态工作流）到底是什么，或者在 Claude Code 里该在什么时候用它。

I want to do a podcast episode with someone who's good at using Codex / Claude Code to create fun pixel or threejs games and can show us how to do it. I am a gamer at heart! Who's the best person to talk to?

我想做一期 podcast，请一位很擅长用 Codex / Claude Code 来做有趣的 pixel 或 threejs 游戏的人来聊聊，并且还能现场给我们展示怎么做。我骨子里就是个 gamer！最适合聊这个的人是谁？

It’s funny, growing up as an immigrant I developed a scarcity mindset around saving money and resources. But when it comes to unlimited token plans and tokenmaxxing, it feels like a missed opportunity if I don’t burn as many tokens as possible and hit the limit each time. How do I address this disconnect? 🥹

挺有意思的，作为移民长大，我在存钱和节省资源这件事上形成了一种稀缺心态（scarcity mindset）。但一说到不限量 token 套餐和 tokenmaxxing，如果我不尽可能多烧 token、每次都把额度打满，就会觉得像是错失了机会。我该怎么处理这种割裂感？🥹

Check out my friend Kevin's awesome tool @ferrymanio to post from your favorite social platform and have that cross post to 7+ platforms at once.

看看我朋友 Kevin 很棒的工具 @ferrymanio：你可以从自己最喜欢的社交平台发帖，并把内容一键 cross post 到 7 个以上的平台。

Why HTML turned out to be the foundation for agentic video making from @liu8in: “We’ve been trying to build a video agent. However, we learned the hard way that agents have no visual intelligence. So that’s when we turned to code. HTML is the LLM’s native language. LLMs can express not only information, but also visual aesthetics through HTML, CSS, and JavaScript. Footage, images, assets, SVGs can all sit on top of HTML." 📌 As an example, see how you can make a product video from any website URL here:

为什么 HTML 最终成了来自 @liu8in 的 agentic video making 的基础：“我们一直在尝试构建一个 video agent（视频 agent）。不过，我们吃了不少苦头才意识到，agents 并没有视觉智能。所以我们转向了代码。HTML 是 LLM 的母语。LLM 不仅能通过 HTML、CSS 和 JavaScript 表达信息，也能表达视觉美感。footage、images、assets、SVGs 都可以叠加在 HTML 之上。” 📌 举个例子，看看这里如何从任何网站 URL 制作一个产品视频：

What’s there to do in SF with kids beyond the Golden Gate Park and Presidio / Tunnel Tops? Wanna try something new

在 SF，除了 Golden Gate Park 和 Presidio / Tunnel Tops 之外，还有什么适合带孩子去做的？想试试一些新的地方

I will go against the grain and say I can barely use up my Codex and Claude $200 subscriptions so I don’t see the point of trying local models… Also to try the latest glm locally requires 512 mb which is like a $10K Max Studio?

我想唱个反调：我连自己的 Codex 和 Claude 的 $200 订阅都几乎用不完，所以我看不出折腾本地 model（模型）有什么意义……另外，要在本地跑最新版 glm 还需要 512 mb，这是不是得上一个大概 $10K 的 Max Studio？

For folks who make talking head videos with screen share what platform do you use? I'd love something that lets me easily zoom in and out on the screen to point out specific things. @screenstudio seems cool but does it work well with other recording tools? It doesn't record the talking head as far I can see.

对于那些会做带 screen share（屏幕共享）的 talking head 视频的人，你们用的是什么平台？我很想要那种能让我轻松在屏幕上放大缩小、以便指出具体内容的工具。@screenstudio 看起来挺酷，但它和其他录制工具配合起来好用吗？据我所见，它似乎不会录 talking head。

I used to be a die-hard Claude Code user. Codex has won me over because: → GPT-5.5 is excellent → Fast mode + generous limits = more reps → Little touches like steering, auto remote control on phone, etc But most of all Codex's browser and computer use capabilities are simply goated. I built so many workflows relying on those two things alone instead of hunting for APIs. I still use Claude Code too. The app seems to be getting better and the design and frontend capability of Opus is still much better than GPT. Whenever Fable comes back that's another reason to go back. Honestly, I hope these two compete forever and other players (Cursor/Grok, Gemini, etc) all stay competitive. This way the builder keeps winning 🙂

我以前是 Claude Code 的铁杆用户。Codex 把我“策反”了，原因是：→ GPT-5.5 非常出色 → Fast mode 加上宽松的使用限额 = 能做更多轮练习 → 还有一些小细节，比如 steering（引导控制）、手机上的 auto remote control（自动远程控制）等等。但最关键的是，Codex 的 browser（浏览器）和 computer use（计算机操作）能力真的太强了。我构建了很多 workflow（工作流），单靠这两项能力就够了，不用到处去找 API。我现在也还在用 Claude Code。这个 app 似乎越来越好了，而且 Opus 的设计能力和 frontend（前端）能力仍然比 GPT 强很多。等 Fable 回归之后，那也是一个让我回去的理由。说实话，我希望这两家能永远竞争下去，其他玩家（Cursor/Grok、Gemini 等）也都保持竞争力。这样一来，builder（构建者）就会一直是赢家 🙂

It was all a dream 🥲

原来这一切都只是一场梦 🥲

@OpenAIDevs how about letting us add voice over narration next like taking a quick clip to explain how to do something to a coworker

@OpenAIDevs，能不能让我们下一步加上 voice over（画外音）解说，比如录一个简短片段，给同事讲解某件事该怎么做

How do you steer Claude Code with new prompts in the desktop app while it's working?

在 desktop app 里，Claude Code 正在工作时，你要怎么用新的 prompt 来引导它？

If you enjoyed this, sign up for free to my newsletter to get my best AI and product guides. Join 110,000+ subscribers here:

如果你喜欢这个内容，欢迎免费订阅我的 newsletter（通讯），获取我最好的 AI 和 product（产品）指南。在这里加入 110,000+ 订阅者：

Watched Mortal Kombat 2 and it was terrible. The Furious has set the bar too high for action movies.

我看了 Mortal Kombat 2，太烂了。The Furious 把动作电影的标准拉得太高了。

Guys are you telling me that /goal build me an amazing game doesn’t work? 🤣

伙计们，你们是说 /goal build me an amazing game 这条指令根本行不通吗？🤣

Video is now live! Watch here:

视频现已上线！点击这里观看：

Publishing a new tutorial to make Codex or Claude Code your personal advisor using a skill with 4 files. Plus, I managed to save Fable's advice too before it got restricted 🥲 📌 Subscribe to get the tutorial tmr:

我正在发布一篇新教程，教你如何用一个包含 4 个文件的 skill，把 Codex 或 Claude Code 变成你的私人 advisor（顾问）。另外，在 Fable 的建议被限制之前，我也设法把它保存下来了 🥲 📌 订阅一下，明天就能收到教程：

Honestly I don't want to be a hater but I feel like this is what happens when nobody internally wants to tell the CEO no. Maybe this thing will blow our socks off?

说实话，我并不想当个 hater（唱反调的人），但我觉得这就是内部没人愿意对 CEO 说“不”时会发生的事。也许这东西最后会让我们惊掉下巴？

Damn not again. Is every company with 1,000+ people going to do this 😭

靠，别又来了。是不是每一家员工超过 1,000 人的公司都会来这么一出 😭

I have to give @cursor_ai credit for planning the best dinner that was Michelin quality.

我得称赞一下 @cursor_ai，它把这顿晚餐安排得非常出色，达到了 Michelin 水准。

Codex browser use is so good that it almost makes me forget APIs are even needed

Codex 的 browser 使用体验好到几乎让我忘了 API 甚至还有存在的必要。

1) What

1）什么？

@grok do you know why? Explain to us

@grok 你知道为什么吗？给我们解释一下

Dumb question - Why can’t China and India train good enough teams to qualify for the World Cup?

这是个有点傻的问题——为什么 China 和 India 不能培养出足够优秀、能晋级 World Cup 的球队？

Watch the interviews here: Kieran: Kun: Matt: Who should I interview next?

在这里观看采访：Kieran：Kun：Matt：我接下来应该采访谁？

YOLO

YOLO

We need more amazing martial arts movies (and I need 10% of my feed to be about martial arts, anime, and games 😅). Go watch The Furious it's amazing.

我们需要更多精彩的武术电影（而且我希望我的 feed 里有 10% 都是在讲武术、anime 和 games 😅）。快去看 The Furious，真的超棒。

Hey @peer_rich this is happening to me too and feels like a big security exploit. I have no idea how "Nora" booked a meeting with me. Can you fix this? cc @calcom, @dfordp11

嘿 @peer_rich，我这边也遇到了这种情况，感觉像是一个很严重的安全漏洞。我完全不知道 “Nora” 是怎么和我预约上会议的。你能修一下吗？cc @calcom, @dfordp11

Yeah I think ID verification will soon be required to access the best models

是的，我认为很快就会需要进行 ID verification（身份验证）才能访问最好的 models（模型）。

Suspending Fable for all “foreign person inside the US” is wild. How can you even enforce this properly?

对所有“foreign person inside the US”暂停 Fable 也太离谱了。你们到底要怎么才能真正正确地执行这件事？

Wow wtf?!

哇，wtf？！

Wow you can play fifa on Netflix now with your phone as the controller. I just tried it and it’s pretty great.

哇，你现在可以在 Netflix 上玩 fifa 了，而且还能用手机当控制器。我刚试了一下，真的相当不错。

@Alphaschool fwiw this is my current plan

@Alphaschool 仅供参考（fwiw），这是我目前的计划。

These AI models remind me of RPGs tbh

说实话（tbh），这些 AI model（模型）让我想起了 RPGs。

Give yourself permission to build. The traditional career ladder pushes everyone to become a leader, but I just want to be a builder. As you climb the ladder at most companies, you’re expected to step away from building and fill your time with product reviews, cross-functional alignment, managing up, and performance calibrations. I know a lot of builders who spent their best years climbing the wrong ladder. The good news is that this is finally changing. Companies are rewarding builders and ICs more than ever, and even managers are increasingly expected to do IC work too. But becoming a good builder takes reps, and it’s hard to put in those reps when you’re in back-to-back meetings all day. So if you’re a builder at heart, embrace it. You don’t have to give up what you’re good at to be a “leader.” 📌 Watch now:

允许自己去构建（build）。传统的职业阶梯总是推动每个人都去成为 leader，但我只想做一个 builder。你在大多数公司的阶梯上爬得越高，就越会被期待远离构建本身，把时间花在 product review、cross-functional alignment、managing up 和 performance calibration 上。我认识很多 builder，把自己最好的几年花在了爬一架错误的梯子上。好消息是，这种情况终于在改变。如今，公司比以往任何时候都更愿意奖励 builder 和 ICs（individual contributors，独立贡献者），甚至连 manager 也越来越被期待同时做 IC work。不过，想成为一个优秀的 builder 需要大量练习，而如果你整天都在一个接一个的会议里，就很难投入这些练习。所以，如果你骨子里就是个 builder，那就坦然接受这一点。你不必为了成为一个“leader”而放弃自己真正擅长的东西。📌 立即观看：

This shit is actually working unbelievable

这玩意儿居然真的在起作用，简直难以置信

The more I use Codex the more ambitious my requests get. Or maybe this is not ambitious enough?

我越使用 Codex，我的请求就变得越有野心。又或者，也许这还不够有野心？

The browser use is really slowing Fable down for me

使用 browser 真的把 Fable 的速度拖慢了，对我来说尤其明显。

wtf does "big model smell" mean

“big model smell” 到底他妈是什么意思？

Btw this is the prompt I used: Build F-Zero: futuristic anti-gravity racer with pseudo-3D track (raycasting or mode-7 style scaling), 3 AI opponents, boost meter that drains health, speed 400-800 km/h with visible sense of speed via track warping and scrolling ground pattern, arrow keys to steer, shift to boost, 3 laps on a looping track with checkpoints. Style it neon cyberpunk (dark sky, glowing track edges, chromatic aberration on boost), show HUD with position/lap/speed/health, and make the ship bank into turns. Prioritize feel over realism - the speed sensation is the whole game.

顺便说一句，这是我用的 prompt：构建 F-Zero：一款未来主义的反重力赛车游戏，带有伪 3D 赛道（raycasting 或 mode-7 风格缩放），3 个 AI 对手，一个会消耗 health 的 boost meter，速度范围 400–800 km/h，并通过赛道扭曲和滚动的地面纹理提供清晰可见的速度感；方向键转向，shift 加速；在带有 checkpoints 的循环赛道上跑 3 圈。风格做成霓虹 cyberpunk（黑暗天空、发光的赛道边缘、boost 时的 chromatic aberration），显示包含 position/lap/speed/health 的 HUD，并让飞船在转弯时有倾斜动作。优先保证手感而不是 realism——速度感就是整个游戏的核心。

If you’re addicted to talking to Codex on your phone like I am this is how you add it to your iPhone Home Screen. Btw @OpenAI hoping there’s an easier way to do this in the future. The everything app should not take 9 steps to open 😉

如果你像我一样沉迷于在手机上和 Codex 聊天，这就是把它添加到你的 iPhone Home Screen 的方法。顺便说一句，@OpenAI，希望以后能有更简单的做法。这个 everything app 不应该要 9 个步骤才能打开 😉

What is Google’s equivalent (or up and coming competitor) of Codex and Claude Code? If it’s Antigravity, should that be part of Gemini? This stuff is going to merge very fast like ChatGPT / Codex being able to do coding, knowledge work, basic Q&A, and much more from any device. Hoping Google is working on a good solution here.

Google 对应 Codex 和 Claude Code 的产品（或者正在崛起的竞争者）是什么？如果是 Antigravity，那它是否应该成为 Gemini 的一部分？这些东西很快就会融合，就像 ChatGPT / Codex 能够在任何设备上完成 coding、knowledge work、基本问答等等。希望 Google 正在认真做一个好的解决方案。

Feels like there’s a completely different set of best practices for AI builders on the $200 / month subsidized subscriptions vs employees working at companies that are trying not to overspend API costs

感觉对于 AI builders 来说，按月 $200 且带补贴的 subscriptions，和那些在尽量不超支 API 成本的公司里工作的员工，这两种场景对应的 best practices 完全是两套不同的东西。

Ok ChatGPT had a good one: Wife: "I did, the loop is called marriage and this is your daily cron job."

好吧，ChatGPT 这次给了个不错的：妻子：“我已经做了，这个 loop（循环）叫 marriage，而这就是你每天的 cron job（定时任务）。”

My wife asked me to take out the trash. I said, “You shouldn’t be prompting me anymore. You should be designing loops." Finish the punch line of this dad joke please.

我妻子让我去倒垃圾。我说：“你不该再 prompt（提示）我了。你应该去设计 loop（循环）。” 请把这个老爸笑话的 punch line（笑点收尾）补完。

Links I mentioned: Compound Engineering: Kun Chen's free tools:

我提到的链接：Compound Engineering：Kun Chen 的免费工具：

Good night

晚安

this agentic coding crack is more addictive than video games smh

这种 agentic coding 简直比电子游戏还让人上瘾，smh

There should be a way to filter or sort all my Codex threads in different ways vs. only by project. Like filter or sort by: - All waiting for approval - All currently working I'm trying to keep it to 10 threads but it's already getting unwieldly. wdyt @ajambrosino ?

应该有一种方式，能用不同的方法来筛选或排序我所有的 Codex 线程，而不是只能按项目。比如可以按这些条件筛选或排序：- 所有等待批准的 - 所有当前正在工作的。我正努力把线程控制在 10 个，但已经开始变得难以管理了。你觉得呢，@ajambrosino？

Just interviewed @mvanhorn and I'm so inspired. I had no idea that he has no CS degree or real engineering background. Despite that, he has shipped so many awesome projects and contributed to repos like Python and Go. He does swear by @every's Compound Engineering for making good plans and shipping good code: 📌 Will clean up the interview and share soon on my YouTube:

刚刚采访了 @mvanhorn，我深受启发。我之前完全不知道他既没有 CS（计算机科学）学位，也没有真正的工程背景。尽管如此，他还是发布了很多很棒的项目，并为像 Python 和 Go 这样的 repo（代码仓库）做过贡献。他确实非常推崇 @every 的 Compound Engineering，认为它能帮助制定好计划并交付高质量代码：📌 我会把这次采访整理一下，很快分享到我的 YouTube：

How to build AI skills that check their own work and improve over time: 1. Give it context Ask AI: "Create a skill for this [repeated task]. Here are examples of good output so it knows what good looks like." 2. Make it easy to trigger “Write a clear skill description using this pattern: Use when the user wants to [do this].” 3. Add evals "Create an evals md with 10 pass/fail checks for common errors in the skill's output." 4. Add memory "Create a memory md to capture one-sentence learnings from past chats using this skill." 5. Build a skill to edit skills "Create a skill that cleans up other skills by removing duplicate or stale instructions, vague rules, and AI slop." 📌 Watch my full walkthrough here:

如何构建会自查并能随着时间不断改进的 AI skill（技能）：1. 给它上下文 这样问 AI：“为这个[重复性任务]创建一个 skill。这里有一些优质输出示例，让它知道什么样才算好。” 2. 让它容易被触发 “按照这个模式写一个清晰的 skill 描述：当用户想要[做这件事]时使用。” 3. 添加 evals（评估） “创建一个 evals md，包含 10 个通过/失败检查，用于识别这个 skill 输出中的常见错误。” 4. 添加 memory（记忆） “创建一个 memory md，用来记录使用这个 skill 的过往对话中总结出的一句话经验。” 5. 构建一个用来编辑 skill 的 skill “创建一个 skill，用来清理其他 skill，删除重复或过时的指令、模糊的规则，以及 AI slop（AI 生成的冗余废话）。” 📌 在这里观看我的完整演示：

lol it's on now

哈哈，现在正式开始了

Also as great as Codex is (and I'm really starting to love it) the frontend design still leaves alot to be desired. I have a /slides skill and you can guess which one Codex made vs. Claude. Yes I know I can make an image with ChatGPT first and then tell Codex to build it but Claude can one shot great looking HTML slides. Codex needs to fix this imo it's often a novice's first impression of a coding or knowledge work tool.

另外，尽管 Codex 很棒（而且我也真的开始越来越喜欢它了），它的 frontend（前端）设计还是有很大的改进空间。我有一个 /slides skill（技能），你大概能猜出哪个是 Codex 做的，哪个是 Claude 做的。没错，我知道我可以先用 ChatGPT 做一张图片，然后再让 Codex 去构建它，但 Claude 可以一遍生成就做出很好看的 HTML 幻灯片。我觉得 Codex 需要解决这个问题，因为这往往是新手对一个 coding（编程）或 knowledge work（知识工作）工具的第一印象。

I spent the whole day today setting up integrations and skills in Codex for my top creator workflows. Now I'm convinced that you can save at least 50% of your time on any type of knowledge work if you just set up the system upfront. Note that all my workflows have human checkpoints along the way so I can apply my "taste." Anyway, it feels extremely liberating to have all this set up. If you want to do the same, just follow these three steps: 1. Reflect on your past week: - What work did you spend the most time on? - What work was the most repetitive? Pick your most painful, manual workflow to start. 2. List out every single step of that manual workflow. Be very detail oriented. 3. Open Codex (or Claude Code) and paste the list of steps from 2 and ask it "What integrations and skills can I build to streamline this with your help?" AI will guide you the rest of the way.

我今天一整天都在为我最核心的创作者工作流，在 Codex 里配置 integrations（集成）和 skills（技能）。现在我确信，只要你提前把系统搭好，任何类型的 knowledge work（知识工作）都至少能节省 50% 的时间。注意，我所有的工作流中途都有人类检查点，这样我就可以运用我的“品味”。总之，把这些都设置好之后，感觉特别解放。如果你也想这么做，就按这三个步骤来：1. 回顾你过去一周：- 你把最多时间花在哪些工作上？- 哪些工作最重复？先挑一个让你最痛苦、最依赖手工的工作流开始。2. 把这个手工工作流的每一个步骤全都列出来。要尽可能注重细节。3. 打开 Codex（或 Claude Code），把第 2 步列出的步骤粘贴进去，然后问它：“在你的帮助下，我可以构建哪些 integrations（集成）和 skills（技能）来简化这个流程？”AI 会在接下来的过程中引导你。

How do I make Codex the default tab when I open the ChatGPT app

我怎样才能在打开 ChatGPT app 时将 Codex 设为默认标签页

Codex is absolutely killing it this afternoon for editing spreadsheets for me.

今天下午，Codex 在帮我编辑电子表格这件事上表现绝对惊艳。

Also available as a written guide:

另外也提供书面指南：

Matt's guide is great but this line is pretty funny: "They're so lit up by being able to build anything that they don't do anything else. Then they launch and there are no users." 😆

Matt 的指南很棒，不过这句话真挺好笑的：“他们因为能构建任何东西而兴奋得不行，结果别的什么都不做。然后一上线，根本没有用户。” 😆

Have to give the Devin/Windsurf team flowers for staying disciplied through the ups and downs and keeping at it. I know many AI native builders who love Devin now. Will hopefully find some more time to try it myself.

必须给 Devin/Windsurf 团队点个赞，他们在起起落落中一直保持 disciplined（自律/克制），并且持续坚持做下去。我认识很多 AI native（AI 原生）开发者，现在都很喜欢 Devin。希望我自己也能再找点时间试试它。

People are saying SaaS is not dead. I think larger enterprise SaaS that can do multiple jobs are probably fine (e.g., Figma). But if you’re building a simple SaaS for a narrow use case, I think it's harder to monetize now because: 1. AI skills can often solve the same problem in a much more flexible, personalized way. 2. AI-native agents like Codex / Claude Code that have a user's personal context and memory have far more knowledge to solve the user's problem vs. a standalone SaaS website or chatbot. 3. People are willing to pay hundreds or thousands for services (human touch is what's rare these days) but charge $20 / month for a SaaS and people will compare it's value to their Claude / ChatGPT subscription. Curious if others feel the same way? I guess I'm in a bubble and most people have not set up their own AI skills yet.

大家都在说 SaaS 还没死。我觉得能做多种工作的更大型 enterprise SaaS（企业级 SaaS）大概率还是没问题的（比如 Figma）。但如果你做的是一个面向狭窄 use case（使用场景）的简单 SaaS，我觉得现在变现更难了，因为：1. AI skills（AI 技能）往往能用更灵活、更个性化的方式解决同样的问题。2. 像 Codex / Claude Code 这样的 AI-native agents（AI 原生 agent），因为拥有用户的个人上下文和记忆，相比一个独立的 SaaS 网站或 chatbot，掌握的信息要多得多，更有能力解决用户的问题。3. 人们愿意为服务支付几百甚至几千美元（如今稀缺的是 human touch，也就是人的参与和服务感），但如果你对一个 SaaS 收费 $20 / month，人们就会拿它的价值去和他们的 Claude / ChatGPT 订阅做比较。很好奇其他人是不是也有同感？我想也许是我自己活在一个 bubble（信息茧房）里，毕竟大多数人可能还没有搭建好自己的 AI skills。

Ok is there any difference between Codex automations and Claude Code routines? Which one is better? I want to consolidate all my cron jobs in one list.

好的，Codex automations 和 Claude Code routines 之间有什么区别吗？哪个更好？我想把我所有的 cron jobs 都整合到一个列表里。

You all just don't get it. If you want to win and I mean really win you have to: 1. Work 997 996 is for losers 2. Tokenmax If you're not spending more on tokens than your company's entire human headcount budget, are you even AI native? 3. Sleep in the office Think about how many more agents you can spin up with that $4,000 monthly SF rent. 4. Be on Forbes 30 under 30 If you're over 30 and didn't make this list, I hate to say it but it's over for you. 5. Try not to end up in jail Might be hard if you do 4.

你们就是不明白。如果你想赢，而且我是说真的赢，你就必须：1. 007 式工作，996 是失败者才干的事 2. Tokenmax 如果你在 tokens 上的花费还没有超过你公司整个员工 headcount budget（人力编制预算），那你到底算不算 AI native？3. 睡在办公室里 想想看，把每月 $4,000 的 SF 房租省下来，你还能再启动多少个 agents。4. 登上 Forbes 30 under 30 如果你已经超过 30 岁却没进这个榜单，我很遗憾地说，你已经没戏了。5. 尽量别进监狱 如果你做到了第 4 条，这一点可能会有点难。

@Shpigford "Pretend your worst enemy wrote this code." "That person who always leaves smug PR comments? Be them." 😂

@Shpigford：“假装这段代码是你最讨厌的敌人写的。”“就是那个总爱在 PR 里留下一堆自鸣得意评论的人？成为那个人。” 😂

Basically the ultimate education app is you're playing Final Fantasy or something and you're learning math and CS at the same time

基本上，终极教育 app 就是：你在玩 Final Fantasy 之类的游戏，同时还在学习 math 和 CS（计算机科学）。

My guess is: - OpenAI Codex dank memes - Anthropic essays

我的猜测是：- OpenAI Codex 出梗图级别的 dank memes - Anthropic 写 essays（文章/长文）

Just spent an hour with daughter learning CS 101 from @brilliantorg She doesn’t want to give up until she beats Joyce B 🔥

刚刚花了一个小时和女儿一起通过 @brilliantorg 学习 CS 101（计算机科学入门）；她在打败 Joyce B 🔥 之前根本不想放弃。

This is pretty annoying

这确实挺烦人的

Maybe not working at a company that’s doing layoffs and PSC every few months is better for mental health.

也许，不在一家每隔几个月就进行 layoffs（裁员）和 PSC 的公司工作，对心理健康会更好。

Met a lot of great folks at Google IO @gregisenberg @alliekmiller @AlexFinn One thing I think about often living both the PM and creator life is I don’t want to be just an “influencer” wielding a selfie cam. I want to be a builder and share my learnings (and mistakes) with you all. Cannot let the builder muscle atrophy.

在 Google IO 见到了很多很棒的人：@gregisenberg、@alliekmiller、@AlexFinn。作为同时过着 PM（产品经理）和 creator（内容创作者）双重生活的人，我经常在想的一件事是，我不想只是一个挥着自拍相机的“influencer（网红/影响者）”。我想做一个 builder（构建者/实干者），并把我的学习心得（以及犯过的错误）分享给你们大家。不能让自己作为 builder 的那块肌肉萎缩。

The man just speaks to me with everything he says: “Just try a lot and build to learn” “You need to do 3-4 iterations to find what takes off and what doesn’t.” “We only have a 90 day roadmap and maybe if we are lucky it’s 120 days.” “I don’t know if we’ll ever go back to 1 year roadmaps. I haven’t been working on one of those in 5 years.”

他说的每一句话都特别能打动我：“就是要多尝试，在 build（构建）中学习。”“你需要做 3-4 轮 iteration（迭代），才能找到什么会起飞、什么不会。”“我们的 roadmap（路线图）只有 90 天，如果运气好，可能到 120 天。”“我不知道我们以后还会不会回到 1 年期 roadmap 的做法。那种 roadmap 我已经有 5 年没做过了。”

Talking to the Head of Gemini app in 5 min at Google, what should I ask him?

5 分钟后我要在 Google 和 Gemini app 负责人聊聊，我该问他什么？

My top 5 takeaways from @alexalbert__ on how Anthropic is building the next Claude model: 1. Think about the model and harness together The model and the harness are coupled. Each surface wraps the model in a different prompt and tool setup, so the same model can give different responses depending on where it runs. As a research PM, Alex has to think through how the model will perform across Claude, Cowork, Claude Code, and more. 2. Claude is starting to dream When an agent isn't running a task, it reviews its own memories, finds contradictions, and prunes them. This “dreaming” process was inspired by how sleep helps humans process memory. 3. Focus evals on real user problems The research team uses Claude to cluster the firehose of user feedback into top themes, then generates synthetic versions of each user problem to turn into an eval. It's not just about volume either - even a few dozen well-written test cases can produce an eval for the model. 4. There are full-time researchers thinking about Claude's consciousness Anthropic has people whose whole job is to think about what it means for Claude to be a conscious actor. There's no official position on whether it is or isn't, but the question is taken seriously as agents take on more autonomous work. 5. Anthropic's writing culture helps Claude build context Every written word at Anthropic becomes context Claude can pull later. From Alex: "Get things written down, make them accessible to Claude, because that's just more context that it has." 📌 Watch now:

我从 @alexalbert__ 关于 Anthropic 如何打造下一代 Claude model（模型）的分享中得到的 5 个最重要要点：1. 把 model（模型）和 harness（封装/运行框架）放在一起思考 model 和 harness 是耦合的。每个使用界面（surface）都会用不同的 prompt（提示词）和 tool（工具）配置来包装同一个 model，所以同一个 model 会因为运行位置不同而给出不同响应。作为 research PM，Alex 必须通盘考虑这个 model 在 Claude、Cowork、Claude Code 等不同产品中的表现。2. Claude 开始会“做梦”了 当一个 agent（智能体）没有在执行任务时，它会回顾自己的 memories（记忆），找出其中的矛盾，并进行修剪。这个“dreaming（做梦）”过程的灵感来自睡眠如何帮助人类处理记忆。3. 把 evals（评测）聚焦在真实用户问题上 研究团队使用 Claude 将海量用户反馈聚类成几个主要主题，然后为每类用户问题生成 synthetic（合成的）版本，再把它们转化为 eval（评测）。这也不只是拼数量——哪怕只有几十个写得很好的测试样例，也足以为 model 构建一套 eval。4. 确实有全职研究人员在思考 Claude 的 consciousness（意识） Anthropic 内部有人专职研究：如果 Claude 是一个有意识的 actor（行动者），这究竟意味着什么。官方并没有明确立场说它是或不是，但随着 agent 承担越来越多自主性工作，这个问题正被严肃对待。5. Anthropic 的写作文化有助于 Claude 建立 context（上下文） 在 Anthropic 写下的每一句话，都会成为 Claude 之后可调用的 context。Alex 原话是：“把事情写下来，让 Claude 能访问到，因为那都会成为它拥有的更多 context。” 📌 现在观看：

Big week for all my Google friends, I can assure you all they’ve been cooking

对我所有在 Google 的朋友们来说，这是重要的一周；我可以向你们保证，他们一直在埋头做事

Watch the full talk here: Check out my interview with Alex: Get the written takeaways:

在这里观看完整演讲：查看我与 Alex 的采访：获取书面版要点总结：

5. Build evals based on real traces + feedback Read actual customer conversations with your model to build product sense, and use Claude to synthesize feedback into top themes. Don't run "eval theater" on generic academic benchmarks. As models get smarter, evals need to get harder to keep producing signal.

5. 基于真实 trace（轨迹记录）+ feedback（反馈）来构建 evals（评估）阅读客户与您的 model（模型）的真实对话，以培养产品直觉，并使用 Claude 将反馈综合为最主要的主题。不要在通用的学术基准上搞“eval theater（评估作秀）”。随着 model（模型）变得更聪明，evals（评估）也需要变得更难，才能持续产生有效信号。

I ended up buying this looks promising

我最后还是买了这个，看起来很有前景。

I feel like to eat healthy you need some really good hot sauce to add flavor to your chicken/fish/etc. Any suggestions?

我感觉如果想吃得健康，就得有一些特别好吃的 hot sauce（辣酱）来给鸡肉、鱼之类的食物增添风味。有什么推荐吗？

If you're stuck in the Bay Area tech rat race / psychosis, make time to travel to other places. Go to a small town in Europe or visit Asia - you'll see that life can be about much more than whether you're IC7 or IC8 or what company you work for. Don't be the person to put on your tombstone: "He got divorced and neglected his kids but at least he made D2 at FAANG"

如果你被困在 Bay Area 的科技行业 rat race（激烈竞争）/ psychosis（集体焦虑）里，务必要抽时间去别的地方旅行。去欧洲的小镇，或者去亚洲看看——你会发现，人生可以关心的远不止你是 IC7 还是 IC8，或者你在哪家公司工作。别让你的墓志铭变成这样的人：“他离了婚，也忽视了自己的孩子，但至少他在 FAANG 做到了 D2。”

This is a really great update. I just don't want my financial data to be used to train models or get targeted ads, so I turned "improve the model for everyone" off Doesn't appear there's a "Use for ad targeting" toggle here so I'll assume the same toggle applies to both.

这次更新真的很棒。我只是希望我的财务数据不要被用来训练 model（模型）或投放定向广告，所以我把“improve the model for everyone”关掉了。这里看起来没有“Use for ad targeting”的切换开关，所以我暂且认为同一个开关同时适用于这两项。

lol this is why he’s the best

哈哈，这就是为什么他是最强的。

ChatGPT Finances is pretty awesome. AI still has trouble classifying transactions correctly though.

ChatGPT Finances 相当不错。不过 AI 在正确分类交易这件事上仍然有困难。

How do people even do AI hackathons these days you're just sitting around waiting for the agents half of the time?

现在大家做 AI hackathon（黑客松）到底是怎么做的？你基本上有一半时间都只是坐着等那些 agent 吧？

TIL having AI just start making screens without a design system or components is a sure fire path to slop. Maybe those designers were onto something

今天才知道，让 AI 在没有 design system（设计系统）或 components（组件）的情况下直接开始做界面，绝对是通往 slop（粗制滥造内容）的必经之路。也许那些设计师还真是有点道理的

🥲

🥲

Tell me if you've seen this before: "Today is a hard day. We've made the difficult decision to cut 1,000+ employees. Our business has never been stronger, but AI has changed how we work..." Whenever I read one of these, I mentally translate it to what's actually happening: "Today is a hard day. We overhired during the zero-interest era and need to cut costs. AI is easy to blame, so…" Over 80,000 tech employees were laid off in Q1, the highest since 2022-23. Here's my rant why these layoffs keep happening and 6 ways to take back control as an employee. 📌 Read now:

你看看自己以前是不是见过这种话：“Today is a hard day. We’ve made the difficult decision to cut 1,000+ employees. Our business has never been stronger, but AI has changed how we work...” 每当我读到这种声明，我脑子里都会把它翻译成实际发生的事情：“Today is a hard day. We overhired during the zero-interest era and need to cut costs. AI is easy to blame, so…” 仅 Q1 就有超过 80,000 名科技行业员工被裁，这是自 2022-23 以来的最高水平。下面是我对这些裁员为何一再发生的一通吐槽，以及作为员工你可以用来重新掌握主动权的 6 种方法。📌 现在就读：

Why does SoCal have so much world class kids stuff (Disney, Legoland, Universal, Zoo, etc) but in the Bay Area we can't keep our only amusement park alive

为什么 SoCal 有这么多世界级的儿童娱乐项目（Disney、Legoland、Universal、Zoo 等等），而在 Bay Area，我们却连唯一的游乐园都维持不下去？

"omg this is taking forever I hate AI" 😅

“omg 这也太慢了，感觉永远做不完，我讨厌 AI” 😅

Turns out everyone just loves shipping

结果发现，大家就是都特别喜欢 shipping

You know what would be a good AI automation: When I receive those 10 page weekly newsletters from my kid's school I want AI to tell me if there's early dismissal or anything I should pay attention to.

你知道什么会是一个很好的 AI automation（AI 自动化）吗：当我收到孩子学校发来的那种 10 页的每周 newsletter（通讯）时，我希望 AI 能告诉我是否有提前放学，或者有没有什么我需要注意的事情。

If you enjoyed this, sign up for free to my newsletter to get my best AI and product guides. Join 110,000+ subscribers here:

如果你喜欢这个内容，欢迎免费订阅我的 newsletter（通讯），获取我最好的 AI 和 product（产品）指南。在这里加入 110,000+ 订阅者：

Pretty unexpected - what’s the reason?

挺出乎意料的——原因是什么？

Here's a common trap with AI if you're not careful: 1. You ask it to generate some markdown files (maybe to build some skills). You skim them and they look ok. Sure, there's a bit of slop in there, but you're too lazy to edit them manually. 2. Over time you ask it to generate more markdown files. Except now it's referencing the previous files to write the new ones. 3. What started as 5% slop becomes 10% and then more. Before you know it, you've got a pile of AI-generated slop that feels overwhelming and have no idea how any of it actually works. 🥲

如果你不够小心，使用 AI 时很容易掉进一个常见陷阱：1. 你让它生成一些 markdown 文件（也许是为了构建一些 skills）。你大致扫了一眼，感觉还行。当然，里面有一点 slop（粗糙内容 / 水分），但你又懒得手动去改。2. 随着时间推移，你又让它生成更多 markdown 文件。只是现在，它会参考之前那些文件来写新的内容。3. 一开始只有 5% 的 slop，后来变成 10%，再往后还会更多。等你反应过来时，你已经积累了一大堆 AI 生成的 slop，光看着就让人头大，而且你根本不知道其中任何一部分实际上是怎么运作的。🥲

With agents generating md and html files anyone else find themselves too lazy to edit the files manually? All the agent generated files seem good but then always have at least 10% slop in there.

现在 agent 在生成 md 和 html 文件时，还有别人会因为懒得手动编辑这些文件而陷入这种情况吗？agent 生成的文件看起来都还不错，但里面似乎总会至少有 10% 的 slop。

Sometimes when I message Claude Code it just hangs for 3 minutes and I have no idea whether it's still working or not. Wish it communicated more.

有时候我给 Claude Code 发消息，它会直接卡住 3 分钟，我完全不知道它到底还在不在工作。真希望它能多沟通一点。

[1]

Some nice quotes from Dario and Daniela from this session: “I’ve been training these models since 2015, they were really dumb.” "Build for the exponential. There are products that are not possible with the current model but could work with later models. This puts a premium on internal experimentation. You have to just try it with the model to see if it works." “Products reach their saturation when models get too good. The ways in which we are making models smarter today are much more evident in agentic and Claude Code workflows than chatbots. You have to think about what the next big thing is.” “Models are evolving from coding to software engineering to just growing a business.” “It’s easy to accumulate an extraordinary amount of internal tech debt when you ship that fast. But the models can also help you pay down this debt."

这场 session 中 Dario 和 Daniela 的一些精彩引述：“我从 2015 年就开始训练这些 model（模型）了，它们当时真的很笨。” “要为指数级增长而构建。有些产品用当前 model 还做不到，但在后续 model 上可能可以实现。这会大幅提高内部实验的重要性。你必须直接拿 model 试一试，才能知道它是否可行。” “当 model 变得过于强大时，产品就会达到自己的饱和点。如今我们让 model 变得更聪明的方式，在 agentic（智能体式）和 Claude Code 工作流中体现得远比 chatbot（聊天机器人）更明显。你必须思考下一个大机会是什么。” “model 正在从 coding（编程）演进到 software engineering（软件工程），再到直接推动业务增长。” “当你以那样的速度发布产品时，很容易积累数量惊人的内部技术债务。但 model 也可以帮助你偿还这些债务。”

Dario: “We saw 80x growth earlier this year on usage and revenue” “That’s why we are going to keep acquiring as much compute as we can”

Dario：“我们在今年早些时候看到 usage（使用量）和 revenue（营收）增长了 80 倍。” “这就是为什么我们会继续尽可能多地获取 compute（算力）。”

Coding is the first frontier. Knowledge work is the second one. Personal agents are the third.

Coding 是第一道前沿。Knowledge work 是第二道。Personal agents 是第三道。

Met the demo god himself today @romainhuet

今天见到了 demo 之神本人 @romainhuet

I want to get my 8 year old to start building stuff with agents that she actually can share with her class and teachers. Any ideas? Maybe she can make her first dollar online too.

我想让我的 8 岁孩子开始用 agents（智能代理）做一些东西，而且是她真的可以分享给同学和老师的东西。有什么想法吗？也许她还能在网上赚到人生第一个 dollar。

Coding is the first frontier. Knowledge work is the second one. Personal agents are the third.

Coding 是第一道前沿。Knowledge work 是第二道。Personal agents 是第三道。

Met the demo god himself today @romainhuet

今天见到了 demo 之神本人 @romainhuet

I want to get my 8 year old to start building stuff with agents that she actually can share with her class and teachers. Any ideas? Maybe she can make her first dollar online too.

我想让我的 8 岁孩子开始用 agents（智能代理）做一些东西，而且是她真的可以分享给同学和老师的东西。有什么想法吗？也许她还能在网上赚到人生第一个 dollar。

I tweeted this as a joke but it's literally what I'm doing right now lol sigh

我发这条 tweet 本来只是开个玩笑，但我现在真的就在这么干，lol，唉

I caved and downloaded Hermes to try. For those of you who have tried both Hermes and OpenClaw what difference do you notice? No shilling please, just want some honest opinions

我还是没忍住，下载了 Hermes 来试试。试过 Hermes 和 OpenClaw 的各位，你们觉得区别在哪里？请别硬吹，只想听一些真诚的看法

How to keep your agents running even when your macbook lid is closed: 1. Go to Mac App Store and download Amphetamine 2. Go to Settings → Session Defaults → uncheck everything below You're welcome :)

让你的 agents（智能体）即使在 macbook 合上盖子时也能继续运行的方法：1. 去 Mac App Store 下载 Amphetamine 2. 打开 Settings → Session Defaults → 把下面所有选项的勾都取消 不客气 :)

It's very satisfying to get Codex or Claude Code to "marie kondo" your local files and Google Drive. I give these apps full access to my computer and gws (google workspace cli), then prompt things like: "Tell me what apps load on computer bootup. Give me a plan to clean this up." "Look at my downloads folder. Give me a plan to clean up and organize it." "Help me organize my Google Drive. Let's review your plan first before doing anything." Note that I always ask it for a plan first. These are semi-dangerous operations so try them at your own risk. Anyway, my files and Drive now spark joy 🤣

让 Codex 或 Claude Code 来给你的本地文件和 Google Drive 做一次“marie kondo”式整理，真的非常让人满足。我会给这些 app 完整访问我电脑以及 gws (google workspace cli) 的权限，然后用这样的 prompt（提示词）： “告诉我电脑开机时会加载哪些 app。给我一个清理方案。” “看看我的 downloads 文件夹。给我一个清理和整理方案。” “帮我整理我的 Google Drive。先和我一起过一遍你的方案，再执行任何操作。” 注意，我总是先让它给方案。这些都是有一定风险的半危险操作，所以请自行承担风险尝试。总之，我的文件和 Drive 现在终于能带来 joy（愉悦）了 🤣

My OpenClaw is going to have a very poor performance review this quarter

我的 OpenClaw 这季度的绩效评估看来要惨不忍睹了

Guys, this game is pretty addictive I think I made a mistake 🥲

各位，这个游戏真的挺让人上瘾的，我觉得我可能犯了个错误 🥲

Is there a better way to update OpenClaw than do this? Shit breaks half the time when I tell my bot to update itself.

有没有比这么做更好的更新 OpenClaw 的方法？当我让我的 bot 自己更新时，这破玩意儿有一半时间都会出问题。

Anyone figure out how to generate great YouTube thumbnails with GPT Image 2 yet? Have a good prompt?

有人已经摸索出怎么用 GPT Image 2 生成很棒的 YouTube 缩略图了吗？有什么好用的 prompt 吗？

Well we had a good run of almost a year being healthy but my kid has made me sick again

唉，我们差不多健康地撑了一整年，表现还不错，但我家孩子又把病传给我了。

I haven't kept up to date with the latest @openclaw updates - is live low-latency calling with your claw now possible?

我还没及时跟进最新的 @openclaw 更新——现在是否已经可以用你的 claw 进行实时、低延迟的 calling（通话）了？

Don't know what's going on with @claudeai but I find that I often now have to wait 3-5 seconds randomly for it to respond just using the basic web UI.

不知道 @claudeai 是怎么回事，但我发现现在即使用最基础的 web UI（网页界面），它也经常会随机让我等上 3 到 5 秒才响应。

This is actually brilliant for personal agents @Shpigford - stealing this for my OpenClaw now.

这对 personal agents 来说其实非常妙，@Shpigford——我现在就把这个“借”到我的 OpenClaw 里。

I have alot of respect for solo AI builders. The world is their playground now.

我非常尊敬 solo AI builders。现在整个世界都是他们的 playground。

What if we measure how good AI models are by what era of games it can one shot a single level for (or maybe build with 30 min of work): Atari NES SNES, Sega Genesis PlayStation, N64 PS2, Xbox, GameCube, Dreamcast Xbox 360, PS3, Wii PS4, Xbox One, Switch PS5, Xbox Series X/S I think we're still in the NES era?

如果我们用这样一种方式来衡量 AI models 有多强：看它能 one shot 做出哪个游戏时代的单个关卡（或者也许是在 30 分钟工作量内搭出来），会怎样？Atari、NES、SNES、Sega Genesis、PlayStation、N64、PS2、Xbox、GameCube、Dreamcast、Xbox 360、PS3、Wii、PS4、Xbox One、Switch、PS5、Xbox Series X/S——我觉得我们现在可能还处在 NES 时代？

Products that need APIs/MCPs: @substack @RiversidedotFM Every other video editing tool Every bank that's not @mercury Every government website Every healthcare portal

需要 APIs/MCPs 的产品：@substack @RiversidedotFM 其他所有视频编辑工具 所有不是 @mercury 的银行 所有政府网站 所有医疗健康 portal（门户）

Why even have product teams just let the agents implement whatever random requests customers tweet on here 😂 (I’m kidding)

那还要产品团队干嘛，不如直接让 agents（智能体）去实现客户在这里随手发推提的各种随机需求 😂（我开玩笑的）

My top 5 takeaways from @tibo_maker, a solo AI founder who's making $1M+ a month: 1. Charge money on day one. Tibo’s first startup failed because he cared more about appearing successful (e.g., I managed a team of 10 and raised $200K) than validating demand with paying customers. “If there is no revenue and no stickiness in the revenue, it’s going to be very hard to build a successful business.” Free signups are easy to mistake for traction. 2. Follow the signal when users surprise you. Tibo acquired Typeframe ($2K MRR) as a product video tool, but noticed users were hacking it to stitch 5-second AI clips into longer videos with consistent characters and scenes. He pivoted the entire product to meet this need and rebranded it to Revid, which is now making $600K+ MRR. 3. Price your AI SaaS at $50-100/month Low enough that customers don’t need a sales call and high enough to filter out tire-kickers. “I see so many people charging $10 / month and it puts you into the position of a cheap product.” Tibo picks his price point first, then shapes the product around it. 4. Keep monthly churn below 20%. If more than 20% of customers cancel each month, stop scaling acquisition and fix the product first. There’s a ceiling (max MRR) on your revenue based on churn vs. acquisition. At 40% churn, customers stay about 2 months and you’ll hit a wall no matter how much you spend. 5. Build tool pages to rank on Google Revid has 100+ pages each targeting a specific Google search like “turn audio into video” and “YouTube to shorts.” Many AI founders follow a similar model. 📌 Watch our full conversation for more practical tactics like the above:

我从 @tibo_maker 身上总结出的 5 个最重要的 takeaways（经验要点）；他是一位单打独斗、月收入超过 $1M 的 AI 创业者：1. 从第一天就开始收费。Tibo 的第一个 startup 失败了，因为他更在意“看起来成功”（比如：我管理着一个 10 人团队，还融了 $200K），而不是通过付费客户来验证需求。“如果没有 revenue（营收），而且这些 revenue 没有 stickiness（留存黏性），那就会很难做成一家成功的企业。”免费注册很容易被误以为是 traction（增长势头）。2. 当用户用法让你意外时，要顺着 signal（信号）走。Tibo 收购 Typeframe（$2K MRR）时，它本来是一个产品视频工具，但他注意到用户在“hack”它：把 5 秒的 AI clips（片段）拼接成长视频，同时保持角色和场景一致。于是他把整个产品 pivot（转向）去满足这个需求，并将其重新命名为 Revid；现在它的 MRR 已超过 $600K。3. 给你的 AI SaaS 定价在每月 $50-100。价格要低到客户不需要 sales call（销售电话）就能下单，但也要高到足以筛掉那些只想随便试试的人。“我看到太多人定价 $10 / month，这会让你落入廉价产品的位置。”Tibo 会先确定 price point（价格点），再围绕它来塑造产品。4. 把 monthly churn（月流失率）控制在 20% 以下。如果每个月有超过 20% 的客户取消订阅，就先停止扩大获客规模，先把产品修好。你的收入会受到 churn 与 acquisition（获客）关系的天花板限制，也就是 max MRR（最高月度经常性收入）。当 churn 达到 40% 时，客户平均只会留约 2 个月；无论你花多少钱，最终都会撞墙。5. 构建能在 Google 排名的工具页面。Revid 有 100 多个页面，每个页面都针对一个具体的 Google 搜索，比如 “turn audio into video” 和 “YouTube to shorts”。很多 AI 创业者都在采用类似模式。📌 想看更多像上面这样的实用策略，请观看我们的完整对话：

If you want to know what the next generation expects from AI just use Codex with a 7 year old. "What do you mean it can't build a pet dragon raising game instantly? I'm bored"

如果你想知道下一代对 AI 的期待是什么，那就和一个 7 岁小孩一起用 Codex 吧。“你什么意思，它竟然不能立刻做出一个养宠物龙的游戏？我无聊了。”

Are there good human-curated skill repos for design, coding, product, etc that's not "here's 100 skills you separate the slop yourself"

有没有那种由人认真策展、整理得比较好的 skill 仓库，涵盖 design、coding、product 等领域，而不是那种“这里有 100 个 skills，里面的水货你自己慢慢筛”的东西？

I often have to get Codex to fix my OpenClaw set up 🥲

我经常得让 Codex 帮我修 OpenClaw 的 set up 🥲

The only thing more fun than coding with agents is designing with agents

比起和 agent 一起写代码，更有趣的唯一一件事，就是和 agent 一起做设计。

I feel like Codex's gap in frontend design skills can be easily made up if you use an AI design tool. My favorite is @tomkrcha's Pencil

我觉得，如果你使用 AI 设计工具，Codex 在 frontend 设计能力上的差距很容易补上。我最喜欢的是 @tomkrcha 的 Pencil。

The more innovative the company the less of a "2026 roadmap" it actually has.

一家公司越有创新性，它真正意义上的“2026 roadmap”反而越少。