Lenny's Podcast: Product | Career | Growth
00:26:10 · Lenny Rachitsky, Claire Vo
[00:26:10] S0: And then you just copy and paste that command. Okay. And then it's off and running, doing stuff, installing, and it's going to ask you some questions. What comes next just as it's loading, just what's going to come next? [00:26:18] S1: Yeah. So it has this really nice onboarding flow, which we will see here. And it's going to ask you, I love this first question, which is, again, I feel like the creator and the maintainers are just such good citizens, and that they say here, this is personal by default. And just so people know what that means, this is not an agent to drop into the group chat. Part of the security posture of OpenClaw is presuming it's working with you and you alone, and you are a trusted sort of instructor of the agent. So we're not going to drop this into your Discord community where anybody and everybody can talk to it. Maybe to your spouse, I have a group chat with my spouse, maybe to a trusted business partner. I have one. ... group chat with my spouse, maybe to a trusted business partner, I have one with my course business partner, but otherwise this is for you and you alone. So you say, "Yes, it's personal use only," and then you do this quick start onboarding and it's going to ask you a couple of things. I won't go through all of it. It's going to ask you what model you want to use, I say use the good models. One, because they're really hardened against prompt injection and some security risks out of the box, and, two, you're just going to get a better experience. So I use a lot of Opus 4.6, Sonnet 4.6 and GPT-5.4.
00:26:18 · Claire Vo, Lenny Rachitsky
[00:26:18] S1: Yeah. So it has this really nice onboarding flow, which we will see here. And it's going to ask you, I love this first question, which is, again, I feel like the creator and the maintainers are just such good citizens, and that they say here, this is personal by default. And just so people know what that means, this is not an agent to drop into the group chat. Part of the security posture of OpenClaw is presuming it's working with you and you alone, and you are a trusted sort of instructor of the agent. So we're not going to drop this into your Discord community where anybody and everybody can talk to it. Maybe to your spouse, I have a group chat with my spouse, maybe to a trusted business partner. I have one. ... group chat with my spouse, maybe to a trusted business partner, I have one with my course business partner, but otherwise this is for you and you alone. So you say, "Yes, it's personal use only," and then you do this quick start onboarding and it's going to ask you a couple of things. I won't go through all of it. It's going to ask you what model you want to use, I say use the good models. One, because they're really hardened against prompt injection and some security risks out of the box, and, two, you're just going to get a better experience. So I use a lot of Opus 4.6, Sonnet 4.6 and GPT-5.4. [00:27:34] S0: And that's versus cheap models is what you're saying there, don't use the cheap stuff?
00:27:34 · Lenny Rachitsky, Claire Vo
[00:27:34] S0: And that's versus cheap models is what you're saying there, don't use the cheap stuff? [00:27:36] S1: Versus cheap models. If you want to go on advanced mode and start to optimize this agent uses this open source model and that agent used that, go for it. I pay for the confidence of the experience and I pay for the security of the nicer model. So it's going to ask you about the model and off provider, and then it's going to ask how you want to talk to it, and it's going to give you a couple options. And I picked Telegram, because I think it's the most beginner-friendly way to set up. Even though you do, and trust me, you'll just have to do it, you do have to talk to this guy called the BotFather on Telegram to set up talking to your OpenClaw, just close your eyes and message the BotFather and don't think too hard about how weird that is. And then once you have the model configured and your channel or how you chat with it configured, you can add a couple tools and then you're off to the races. [00:28:33] S0: And once you get that initial stuff going, you can connect it to WhatsApp, you can connect it to iMessage, something that I did, which is a little complicated but easy to do, and then you could talk to it over email, Slack. [00:28:44] S1: Yeah, anything you want. [00:28:46] S0: Okay, cool. So, I know one of the steps is it asks you just like, "What am I? What do you want me to do? Who do you want me to be?" Talk about that step.
00:28:46 · Lenny Rachitsky, Claire Vo
[00:28:46] S0: Okay, cool. So, I know one of the steps is it asks you just like, "What am I? What do you want me to do? Who do you want me to be?" Talk about that step. [00:28:53] S1: Let's do that. I'm going to pull up this screen, which is a agent I just configured before the show. Its name is Q. And Q is going to be my kid's agent. I have a personal assistant for the family, I have a personal assistant for work, my kids have too much homework or they think they have too much homework, and so I'm going to set up Q. [00:29:17] S0: This is an awesome idea, by the way. Just the idea of giving... Versus ChatGPT.
00:29:17 · Lenny Rachitsky, Claire Vo
[00:29:17] S0: This is an awesome idea, by the way. Just the idea of giving... Versus ChatGPT. [00:29:20] S1: Jesse Genet has inspired me on all things OpenClaw and kids, so it's a great idea. I haven't said anything to Q yet, so we're just going to see what he says, so I'm going to say hi. And I'm running OpenClaw right here in my terminal just because it's easier to show for the screen share, but you can just do this in Telegram as well. And it says, "Hey, I just came online. Who am I? Who are you?" And so, again, product people, designers pay attention, if you think about the onboarding experience of an agent, you know what I do at ChatPRD, I force you to press these buttons and tell me who you are and write a bio and all this stuff in these structured fields, and I think that's such an old mental model of how to design an onboarding experience, and this is a very interesting one where it just says like, "Who am I and who are you? And we're going to figure this out together." And so I'm going to say, "You're Q. You are a elementary school teacher and ex professor, slash, scientist who is going to help me and my kids with their academic and extracurricular pursuits." I'm clear. This is a trick, I call this the brain transplant. So because Q is living on the same machine as my other agents, I say, "You can look in my other agent, Polly's Soul..." [00:30:52] S0: Just an amazing thing to write.
00:30:52 · Lenny Rachitsky, Claire Vo
[00:30:52] S0: Just an amazing thing to write. [00:30:53] S1: "... to learn more about me. Finn also has some good info about me and the kids." Okay, so this is, I think, an advanced trick, but because all these agents live in the same folder and workspace, same computer, if you give them permission, you can also lock them down so they know that, they can actually go look in other places. Oh no, he's sandboxed. Well, okay. Well, we'll skip this. OpenClaw has done a good job of locking things down by default and forcing you to open them up, which I appreciate. [00:31:30] S0: You just can't look at the other soul, is that what's going on there? [00:31:32] S1: Yeah, I just can't look at- [00:31:32] S0: Security for the win.