跳到主内容
@wquguru
精选85Lex Fridman(YouTube)产品发布/更新

OpenClaw 创始人 Peter Steinberger 访谈:AI 代理的崛起

OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491

原文
发到 X
推荐理由

做 AI 代理的同学必看,OpenClaw 代表了从语言到行动的范式转变,开源且社区活跃,值得深入研究其架构和安全性设计。

  • I watched my agent happily click the "I'm not a robot" button. I made the agent very aware. Like, it knows what his source code is. It understands th- how it sits and runs in its own harness. It knows where documentation is. It knows which model it runs. It understands its own system that made it very easy for an agent to... Oh, you don't like anything? You just prompted it to existence, and then the agent would just modify its own software.

People talk about self-modifying software, I just built it. I actually think wipe coding is a slur. - You prefer agentic engineering? - Yeah, I always tell people I'd- I do agentic engineering, and then maybe after 3:00 AM, I switch to wipe coding, and then I have regrets on the next day. - What a walk of shame. - Yeah, you just have to clean up and, like, fix your sh- shit. - We've all been there. - I used to write really long prompts.

And by writing, I mean, I don't write, I- I- I talk, you know? These- these hands are, like, too- too precious for writing now. I just- I just use bespoke prompts to build my software. - So, you, for real, with all those terminals, are using voice? - Yeah. I used to do it very extensively, to the point where there was a period where I lost my voice. - I mean, I have to ask you, just curious. I- I know you've probably gotten huge offers from major companies.

Can you speak to who you're considering working with? - Yeah. - The following is a conversation with Peter Steinberger, creator of OpenClaw, formerly known as MoldBot, ClawedBot, Clawdus, Claude, spelled with a W as in lobster claw. Not to be confused with Claud, the AI model from Anthropic, spelled with a U. In fact, this confusion is the reason Anthropic kindly asked Peter to change the name to OpenClaw. So, what is OpenClaw?

It's an open-source AI agent that has taken over the tech world in a matter of days, exploding in popularity, reaching over 180,000 stars on GitHub, and spawning the social network mold book, where AI agents post manifestos and debate consciousness, creating a mix of excitement and fear in the general public. And a kind of AI psychosis, a mix of clickbait fearmongering and genuine, fully justifiable concern about the role of AI in our digital, interconnected human world.

OpenClaw, as its tagline states, is the AI that actually does things. It's an autonomous AI assistant that lives in your computer, has access to all of your stuff, if you let it, talks to you through Telegram, WhatsApp, Signal, iMessage, and whatever else messaging client. Uses whatever AI model you like, including Claude Opus 4.6 and GPT 5.3 Codex, all to do stuff for you. Many people are calling this one of the biggest moments in the recent history of AI, since the launch of ChatGPT in November 2022.

The ingredients for this kind of AI agent were all there, but putting it all together in a system that definitively takes a step forward over the line from language to agency, from ideas to actions, in a way that created a useful assistant that feels like one who gets you and learns from you, in an open source, community-driven way, is the reason OpenClaw took the internet by storm. Its power, in large part, comes from the fact that you can give it access to all of your stuff and give it permission to do anything with that stuff in order to be useful to you.

This is very powerful, but it is also dangerous. OpenClaw represents freedom, but with freedom comes responsibility. With it, you can own and have control over your data, but precisely because you have this control, you also have the responsibility to protect it from cybersecurity threats of various kinds. There are great ways to protect yourself, but the threats and vulnerabilities are out there. Again, a powerful AI agent with system- level access is a security minefield, but it also represents the future.

Because when done well and securely, it can be extremely useful to each of us humans as a personal assistant. We discuss all of this with Peter, and also discuss his big-picture programming and entrepreneurship life story, which I think is truly inspiring. He spent 13 years building PSPDF Kit, which is a software used on a billion devices. He sold it, and for a brief time, fell out of love with programming, vanished for three years, and then came back, rediscovered his love for programming, and built, in a very short time, an open source AI agent that took the internet by storm.

He is, in many ways, the symbol of the AI revolution happening in the programming world. There was the ChatGPT moment in 2022, the DeepSeek moment in 2025, and now, in '26, we're living through the OpenClaw moment, the age of the lobster. The start of the agentic AI revolution. What a time to be alive. This is a Lex Fridman podcast. To support it, please check out our sponsors in the description, or you can also find links to contact me, ask questions, give feedback, and so on.

And now, dear friends, here's Peter Steinberger. The one and only, the Clawed Father. Actually, Benjamin predicted it in his tweet. "The following is a conversation with Claude, a respected crustacean." It's a hilarious-looking picture of a lobster in a suit, so I think the prophecy has been fulfilled. Let's go to this moment when you built a prototype in one hour, that was the early version of OpenClaw. I think this, Story's really inspiring to a lot of people because this prototype led to something that just took the internet by storm.... and became the fastest-growing repository in GitHub history, with now over 175,000 stars.

So, what was the story of the one-hour prototype? - You know, I wanted that since April. - A personal assistant. AI personal assistant. - Yeah. And I, I played around with some other things, like even stuff that gets all my WhatsApp, and I could just run queries on it. That was back when we had GPT-4.1, with the one million context window. And I, I pulled in all the data and then just asked him questions like, "What makes this friendship meaningful?" - Mm-hmm. - And I got some, some really profound results.

Like, I sent it to my friends and they got, like, teary eyes. - So, there's something there. - Yeah. But then I... I thought all the labs will, will, will work on that. So I, I moved on to other things, and that was still very much in my early days of experimenting and pl- playing. You know, you have to... That's how you learn. You just like, you do stuff and you play. And time flew by and it was November. I wanted to make sure that the thing I started is actually happening.

I was annoyed that it didn't exist, so I just prompted it into existence. - I mean, that's the beginning of the hero's journey of the entrepreneur, right? And you've even with your original story with PS PDF kit, it's like, "Why does this not exist? Let me build it." And again, here's diff- a whole different realm, but similar maybe spirit. - Yeah, so I had this problem. I tried to show PDF on an iPad, which should not be hard. - This is like 15 years ago, something like that. - Yeah.

Like the most, the most random thing ever. And suddenly, I had this problem and I, I wanted to help a friend. And there was, there was... Well, not like nothing existed, but it was just not good. And like... Like I tried it and it was like very, "Nah." Like, "Hmm, I can do this better." - By the way, for people who don't know, this led to the development of PS PDF kit that's used on a billion devices. So, the... It turns out that it's pretty useful to be able to open a PDF. - You could also make the joke that I'm really bad at naming. - Yeah. - Like, name number five on the current project.

And even PS PDF doesn't really roll from the tongue. - Anyway, so you said "Screw it. Why don't I do it?" So what was the... What was the prototype? What was the thing that you... What was the magical thing that you built in a short amount of time that you were like, "This might actually work as an agent," where I talk to it and it does things? - There was... Like,

原文超出正文长度上限,此处截断——上游还有内容,完整版见上方「原文 ↗」。

更进一步:量化金融体系

看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力

进入量化体系 →

相似阅读

另一事件,读法相近