This is what production-grade visual AI looks like.
这就是达到 production-grade（生产级）标准的 visual AI 的样子。
AA
Aditya Agarwal
@adityaag ↗General Partner @SouthPkCommons, Co-Founder @Bevel_Health | Ex: Early Eng @facebook, CTO @Dropbox, Board @Flipkart | Optimist, Builder, Dad
2 最新33 累计22 期
每条推文 hover 显示单独 ▶
Things are only impossible to build until someone builds them.
一件事在被人造出来之前，看起来都只是“不可能做到”。
As an investor, these are the moments that remind you why you fund startups to do important things.
作为一名投资者，这样的时刻会提醒你：你之所以投资 startups（初创公司），是为了让它们去做重要的事情。
I have had the chance to go through two IPOs (Meta and Dropbox). Fabulous wealth tends to amplify deeper desires, not create new ones. The mainstream narrative is: "Early Employees make lots of money and buy mansions and go to the beach". For some: Yes, they want to chill. That is great. I am happy for them. But for a lot of folks, it is a chance to do even crazier and wackier stuff. Start new things. Fund new things. Keep the crazy loop of Silicon Valley going. It's going to be awesome to have a lot of folks with liquidity in the coming months. Onwards!.
我有机会经历过两次 IPO（Meta 和 Dropbox）。巨额财富往往会放大一个人更深层的欲望，而不是创造出新的欲望。主流叙事通常是：“早期员工赚了很多钱，然后买豪宅、去海边度假。” 对有些人来说：是的，他们就是想放松一下。这很好。我也为他们高兴。但对很多人来说，这反而是一个去做更疯狂、更离奇事情的机会。创办新东西。资助新东西。让 Silicon Valley 的这套疯狂循环继续转下去。在接下来的几个月里，会有很多人拥有 liquidity（流动性），这将会非常棒。继续前进！
[1]
Sometimes speed is just impatience disguised as ambition.
有时候，速度不过是把不耐烦伪装成了雄心。
A lot of roles will now have engineering infused into them. Marketing Engineer is a great example of this.
现在，很多岗位都会融入 engineering（工程）能力。Marketing Engineer 就是一个很好的例子。
Very excited to welcome Harshit Madan and Rohan Choudhary to the team!
非常高兴欢迎 Harshit Madan 和 Rohan Choudhary 加入团队！
Schrep and the Gigascale team are building something special. They have been incredible partners to @southpkcommons portfolio companies. @schrep is someone who I have looked up to for a long time and I am very excited to continue to collaborate with him.
Schrep 和 Gigascale 团队正在打造一些特别的东西。他们一直是 @southpkcommons 被投公司极其出色的合作伙伴。@schrep 是我长期以来一直很敬佩的人，我也非常期待继续与他合作。
Most people I talk to in India have 1 child. Some have 2. When I tell them I have three, their first reaction is total surprise followed by "well, accidents happen". The main reason for one? Raising kids is just really hard, especially for nuclear families in cities.
我在 India 接触到的大多数人都只有 1 个孩子，也有些人有 2 个。当我告诉他们我有三个时，他们的第一反应是完全吃惊，接着就会说一句“唉，意外总会发生”。只生一个的主要原因是什么？养孩子实在太难了，尤其是对城市里的 nuclear families（核心家庭）来说。
Legendary run my dear friend. Proud to have played a part in the journey. I cannot wait to see what you do next.
了不起的一段历程，我亲爱的朋友。很自豪能在这段旅程中贡献一份力量。我已经迫不及待想看看你接下来会做什么了。
Building at the frontier? Exploring what's worth building? Both are good reasons to apply to @southpkcommons.
正在前沿领域进行构建？在探索什么值得去构建？这两点都是申请 @southpkcommons 的充分理由。
It was amazing to have Feross do his -1 journey @southpkcommons and to have been an early partner. Very proud of the progress the company has made. Lots to build!
Feross 能在 @southpkcommons 完成他的 -1 旅程，而且我们能成为早期合作伙伴，真的很了不起。公司取得的进展让我非常自豪。还有很多东西要继续建设！
Apply to @southpkcommons if you'd like to join us
如果你想加入我们，就申请 @southpkcommons 吧。
The future is bright. We will have amazing things at the intersection of AI+Atoms.
未来一片光明。我们将在 AI+Atoms 的交汇处迎来令人惊叹的事物。
On Friday I used AI to write a lot of code. And then on Saturday I watched humans play sports (FA Cup Final) and dance (my kid's dance recital). I am not worried that we won't have meaning and purpose.
周五我用 AI 写了很多代码。然后周六我看人类进行体育比赛（FA Cup Final）和跳舞（我家孩子的舞蹈汇演）。我并不担心我们会失去意义和目标。
[1]
This is such an insane POV. @AOC if you join @southpkcommons we can help you find your life’s work and also make you a billion dollars. Seriously, you can just do things.
这真是一个离谱到爆的 POV（观点）。@AOC，如果你加入 @southpkcommons，我们可以帮你找到你一生真正要做的事业，还能让你赚到 10 亿美元。说真的，你完全可以直接去做事情。
Apply to join the Embodied AI Hackathon at @southpkcommons by May 12th ⬇️
在 5 月 12 日前申请加入位于 @southpkcommons 的 Embodied AI Hackathon ⬇️
We have an incredible group of hard tech founders at @southpkcommons right now. Come compete with them. Embodied AI Hackathon, SF, May 15-17.
现在 @southpkcommons 聚集了一批非常厉害的 hard tech 创始人。来和他们同场竞争吧。Embodied AI Hackathon，SF，5 月 15 日至 17 日。
We have never desired to be an accelerator. Velocity is not interesting without a grounding towards true north.
我们从未想过要成为一个 accelerator（加速器）。如果没有朝向 true north（真正北方、根本目标）的稳固锚定，velocity（速度）就没有意义。
We have never desired to be an accelerator. Velocity is not interesting without a grounding towards true north.
我们从未想过要成为一个 accelerator（加速器）。如果没有朝向 true north（真正北方、根本目标）的稳固锚定，velocity（速度）就没有意义。
If you get out of the console and terminal, you will start to see that there is a lot of hard tech being built in America. It is deeply deeply inspiring and exciting. Let’s shoot things into space ✅
如果你走出 console 和 terminal，你就会开始看到，美国正在打造大量真正硬核的 tech（技术）。这非常、非常鼓舞人心，也令人兴奋。让我们把东西发射到太空去吧 ✅
We’re going through a critical inflection point for cybersecurity in real time.
我们正在实时经历 cybersecurity（网络安全）的一个关键拐点。
The attackers have AI. So do the defenders. Cybersecurity has a new shape: AI is both the threat and the only viable solution. @nikesharora, CEO of @PaloAltoNtwks, joined me at @southpkcommons to break down what’s next. (00:30) Big problems vs. fast wins (04:00) Joining Google (07:00) Larry Page's product obsession (10:00) He read every hiring packet himself (13:00) What Silicon Valley gets wrong about Masa Son (16:00) Successful founders never wish they took less risk (17:00) Joining Palo Alto knowing nothing about cybersecurity (20:00) What incumbents got wrong when ChatGPT launched (23:00) Security was never built into AI (28:00) No enterprise knows what's running inside its stack (31:00) AI finds bad code faster than humans ever could (35:00) The only way to fix the chaos is more AI (38:00) AI won't just automate work—it raises the floor (41:00) Foundation models vs. specialized stacks (45:00) Why communication is 30% of the job at scale
攻击者拥有 AI。防御者也一样。cybersecurity（网络安全）呈现出一种新形态：AI 既是威胁，也是唯一可行的解决方案。@nikesharora，@PaloAltoNtwks 的 CEO，与我一同在 @southpkcommons 讨论了接下来会发生什么。（00:30）重大问题 vs. 快速见效的成果（04:00）加入 Google（07:00）Larry Page 对产品的痴迷（10:00）他亲自阅读每一份招聘材料（13:00）Silicon Valley 对 Masa Son 的误解（16:00）成功的创始人从不会希望自己当初承担更少风险（17:00）在对 cybersecurity 一无所知的情况下加入 Palo Alto（20:00）ChatGPT 发布时，老牌 incumbents（既有企业）犯了什么错（23:00）安全性从未被内建到 AI 中（28:00）没有哪家企业真正知道自己的技术栈内部在运行什么（31:00）AI 发现糟糕代码的速度比人类快得多（35:00）解决这种混乱的唯一方法，是更多 AI（38:00）AI 不只会自动化工作——它还会抬高最低水平（41:00）foundation models（基础模型）vs. specialized stacks（专用技术栈）（45:00）为什么在规模化之后，沟通占工作内容的 30%
Some observations on agents from a lot of time spent installing, using, debugging and building things with OpenClaw, Hermes, etc. - Most people who rave about OpenClaw, HermesAgent etc. have not actually used it. Facts. - Agents are developer products today. They are so so so far from being consumer grade in terms of delight, simplicity, reliability etc. - I wish the agents generated more dynamic interfaces. Don't give me chat. Give me interactive, dynamic apps/webpages. I suspect this will be a next big vector of innovation. - The agents are ultimately a wrapper to code-generation and tool-calling....not to "generative AI". This is important to grok. These agents are doing most of their "work" through iterative tool calling. - Once you understand the core loop behind agentic AI, old school "chat AI" doesn't make sense. The idea of an "always-on" agent that actually adapts, learns and maintains state is a LOT more compelling than a one-off chat prompt. The latter feels much more like a search query. - So much of mangling with agents today is basically setting it up to use browsers and/or your computer. And there are so many constraints! - Installing and running openclaw feels a lot like running Linux in the early 2000s. You spend a lot of time compiling device drivers, setting up configurations etc. You feel a real sense of accomplishment from jerry-rigging everything together but it is a LOT of work to get to the point of utility
关于 agent 的一些观察，来自我花了大量时间安装、使用、debug（调试）以及用 OpenClaw、Hermes 等搭东西的经验。—— 大多数对 OpenClaw、HermesAgent 等赞不绝口的人，其实并没有真正用过它。事实如此。—— 如今的 agent 还是开发者产品。无论在愉悦感、简洁性、可靠性等方面，它们距离 consumer grade（消费级）都还差得非常非常远。—— 我希望这些 agent 能生成更多动态界面。别只给我 chat（聊天界面），给我可交互、动态的 app / webpage（应用 / 网页）。我怀疑这会是下一个重要的创新方向。—— 这些 agent 归根结底是 code-generation（代码生成）和 tool-calling（工具调用）的 wrapper（封装）……而不是“generative AI（生成式 AI）”本身。理解这一点很重要。这些 agent 的大部分“工作”，实际上都是通过迭代式工具调用完成的。—— 一旦你理解了 agentic AI（智能体式 AI）背后的核心循环，老派的“chat AI”就显得不太说得通了。相比一次性的聊天 prompt（提示词），那种“always-on（始终在线）”、能够真正适应、学习并维持 state（状态）的 agent，要有吸引力得多。后者更像是一条搜索查询。—— 如今折腾 agent 的很大一部分，本质上就是把它配置好，让它能使用浏览器和/或你的电脑。而且限制多得惊人！—— 安装并运行 openclaw 的感觉，很像在 2000 年代初运行 Linux。你会花很多时间编译设备驱动、设置各种配置等。把一切东拼西凑起来时，你会有一种真正的成就感，但要走到“真正有用”那一步，工作量实在太大了。
Bevel 3.0 The point was never to be a tracking and analytics company. Each release of @bevel brings us closer to our long-term goal: To be the most intelligent health and biology companion that truly understands you and can help you lead a healthier and happier life. It was to be intelligent enough to help you truly be healthy. We are taking another step towards that with Bevel 3.0.
Bevel 3.0 的目标从来都不是成为一家 tracking 和 analytics 公司。@bevel 的每一次发布，都让我们离长期目标更近一步：成为最智能的健康与生物学 companion（陪伴助手），真正理解你，并帮助你过上更健康、更快乐的生活。它的目标，是足够智能，能够真正帮助你保持健康。通过 Bevel 3.0，我们正朝这个方向再迈进一步。
[1]
Before @RamaswmySridhar was CEO of @Snowflake, he was the guy who scaled Google's ads business from $1.5B to $ 100B. He's coming by @southpkcommons next week. Join us.
在 @RamaswmySridhar 成为 @Snowflake 的 CEO 之前，他就是那个把 Google 的 ads 业务从 15 亿美元扩张到 1000 亿美元的人。他将于下周来到 @southpkcommons。欢迎加入我们。