How the OpenClaw foundation bullet-proofed its future (w/Dave Morin) | E2257

Summary

Dave Morin joins the show to explain how the OpenClaw Foundation came into existence and what its goals are. While Peter Steinberger retains directional control of the open-source project, sustaining it over time requires a corporate entity, well-heeled sponsors, and dedicated maintainers. Morin details how he got “Claw-Pilled,” the foundation’s governance structure, the economics of open-source maintainers, and his physical OpenClaw setup at home.

The episode also puts the day’s critical AI stories to Morin: the Anthropic-DoW controversy, whether OpenAI and Anthropic are catching up to OpenClaw in agentic capabilities, and Apple’s slow movement in AI. Two demos round out the show — Runtools’ CTO Greg Kara shows a 3D interface for OpenClaw that lets users watch their agents work in real time, and PickleWatch founder George Yameen demonstrates an agentic workflow comparing OpenClaw with Claude’s Chrome plugin for cost-effectiveness. Jason closes with advice for founders about stumbling onto new ideas and his take on the Department of Welfare’s handling of the Anthropic situation.

Highlights

How Dave Morin Got Claw-Pilled

“How Dave Morin got Claw-Pilled — from skeptic to foundation builder.” — Dave Morin, 1:58

yt-dlp --download-sections "*01:58-03:54" "https://www.youtube.com/watch?v=KEHMoI00q3w" --force-keyframes-at-cuts --merge-output-format mp4 -o "dave-morin-claw-pilled.mp4"

The OpenClaw Foundation Structure

“How long has the OpenClaw Foundation existed? How much authority does Peter Steinberger retain?” — Jason Calacanis, 13:39

yt-dlp --download-sections "*13:39-17:54" "https://www.youtube.com/watch?v=KEHMoI00q3w" --force-keyframes-at-cuts --merge-output-format mp4 -o "openclaw-foundation-governance.mp4"

Are Anthropic and OpenAI Catching Up?

“Are Anthropic and OpenAI catching up to OpenClaw in agentic capabilities?” — Jason Calacanis, 43:46

yt-dlp --download-sections "*43:46-48:22" "https://www.youtube.com/watch?v=KEHMoI00q3w" --force-keyframes-at-cuts --merge-output-format mp4 -o "anthropic-openai-catching-up.mp4"

Runtools 3D OpenClaw Environment

“Runtools.AI demo of 3D OpenClaw environment — watch your agents work in real time.” — Greg Kara, 52:38

yt-dlp --download-sections "*52:38-55:45" "https://www.youtube.com/watch?v=KEHMoI00q3w" --force-keyframes-at-cuts --merge-output-format mp4 -o "runtools-3d-openclaw.mp4"

Apple’s Slow Movement in AI

“Apple’s slow movement and hesitance to deprecate products in the age of AI.” — Dave Morin, 32:42

yt-dlp --download-sections "*32:42-39:13" "https://www.youtube.com/watch?v=KEHMoI00q3w" --force-keyframes-at-cuts --merge-output-format mp4 -o "apple-slow-ai-movement.mp4"

Key Points

• Dave Morin gets Claw-Pilled (1:58) - How Dave discovered and became passionate about OpenClaw
• Jason’s critical moments (3:54) - Jason’s key moments that convinced him OpenClaw was transformative
• Dave’s OpenClaw skills (7:07) - How Dave showcases and uses OpenClaw capabilities
• Foundation timeline (13:39) - When the OpenClaw Foundation was established
• Peter’s authority (16:20) - How much control Peter Steinberger retains over the project’s direction
• Air traffic control model (17:54) - How coordination works for a project like OpenClaw
• Open-source economics (21:59) - The economics of funding and sustaining open-source maintainers
• Dave’s physical setup (24:07) - Dave Morin’s home OpenClaw hardware configuration
• Jason’s OpenClaw uses (27:16) - How Jason uses OpenClaw to curate reading, fashion, and travel
• Second wave ideas (31:11) - Why the second wave of ideas gets most interesting
• Apple’s hesitance (32:42) - Apple’s slow movement and reluctance to deprecate products
• OpenClaw coaching (39:13) - How Jason uses OpenClaw to coach the next generation of executives
• Competition assessment (43:46) - Whether Anthropic and OpenAI are catching up to OpenClaw
• Runtools 3D demo (52:38) - 3D interface for watching OpenClaw agents work in real time
• PickleWatch demo (56:10) - Comparing OpenClaw with Claude Chrome plugin for cost-effectiveness
• DoW-Anthropic controversy (1:05:53) - Jason’s take on the Department of Welfare’s handling of Anthropic
• Funding Anthropic alternatives (1:10:20) - Would Jason fund the tech that Anthropic won’t supply?

Mentions

Companies

• OpenClaw Foundation (13:39) - New foundation to sustain and govern the OpenClaw project
• Anthropic (43:46) - AI company discussed for catching up to OpenClaw and DoW controversy
• OpenAI (43:46) - Discussed for agentic AI competition with OpenClaw
• Apple (32:42) - Criticized for slow AI adoption and hesitance to deprecate products
• Runtools.AI (48:22) - Greg Kara’s 3D OpenClaw visualization tool
• PickleWatch (55:45) - George Yameen’s pickleball analytics company
• Uber AI Solutions (9:40) - Sponsor, providing real-world AI solutions

Products & Technologies

• OpenClaw (0:00) - Open-source AI agent framework at center of the episode
• Claude Chrome Plugin (56:10) - Compared with OpenClaw for cost-effectiveness in demo

People

• Dave Morin (0:00) - Key figure in OpenClaw Foundation establishment
• Peter Steinberger (16:20) - Creator and directional leader of OpenClaw
• Greg Kara (48:22) - CTO of Runtools.AI
• George Yameen (55:45) - Founder of PickleWatch

Surprising Quotes

“The second wave of ideas is where things get the most interesting — the obvious applications come first, but the truly transformative ones follow.” — Dave Morin, 31:11

“Jason uses OpenClaw to coach the next generation of executives — it’s like having a tireless mentor that never sleeps.” — Jason Calacanis, 39:13

“Would Jason fund the tech that Anthropic won’t supply? Absolutely.” — Jason Calacanis, 1:10:20

Transcript

Jason Calacanis: 0:00 Hey everybody welcome back to TWIST. March 2nd, 2026. AO39 in the year of our lord the Open Claw. 39 days after we first talked about Open Claw here on the pod and cackling in the background, my old friend, a true bestie, one of the great investors, human beings, entrepreneurs, commentators, podcasters of our generation here, Dave Morin. How are you doing brother?

Dave Morin: 0:28 Good to be here Jason, thanks for having me.

Jason Calacanis: 0:30 Oh, it’s so great to have you. Of course my co-host Alex is here. It’s been a minute, Alex is here and he’s ready to rock and roll. Yep.

Lon Harris: 0:41 Yes sir.

Dave Morin: 0:42 Alex with the best podcasting voice.

Jason Calacanis: 0:44 Oh, he’s great.

Lon Harris: 0:46 You missed that. Dave was giving me props before the show. It was the best day of my life.

Jason Calacanis: 0:50 You know some people are gifted with the broadcast voice, others of us you know not so much so we have to make it work.

Dave Morin: 0:57 We make it work.

Jason Calacanis: 0:59 We make it work with the limited resources we have. Dave, I wanted to have you on because you too have been Claw-pilled. You first, and then of course when Open Claw was aqua-hired slash we’ll get into whatever the transaction was and our friend Sam Altman bought it sort of, they said big concerns, open source project is going to get the plug pulled on it, this is a disaster, everybody has a moment of panic and then it’s announced that you will become the first board member of the Open Claw Foundation and everybody took a deep sigh of relief because putting you in there was obviously the signal that this is going to be done right and done fairly because that’s your reputation. Tell us, two questions. Let’s start with how did you get Claw-pilled and why is this technology so inspiring that it went to the top of GitHub today even beating React, which was another one of those cult-like phenomenons in the open source community?

Dave Morin: 1:59 Yeah, late December I was in a couple of group chats that go back to like the early 2000s, like when you and I met.

Jason Calacanis: 2:09 Web 2.0.

Dave Morin: 2:11 Yeah, it’s the old Web 2.0, you know, just guys that are always tinkering around, like funny enough it’s like people from the old Foo Camp days and things like that.

Jason Calacanis: 2:20 Oh wow, there’s an old reference.

Dave Morin: 2:23 Yeah. And one of them said, you know, ‘Hey, has anyone seen this Open Claw thing?’ It was called Clawbot back then. And so I gave it a try. This was like the late December, early January. And within the first 24 hours it was super clear to me that this was different and the future in a way that I hadn’t really experienced since maybe the early 2000s. Like I got genuinely excited about building things with this. And it was, I think I had sent a tweet at the time, I was like, ‘this is the first time I felt like I’ve been living in the future since ChatGPT.’

Greg Kara: 3:00 GPT and and uh in my mind I was like this is like iPhone this might even be like LAMP stack like this is like this is like Linux level interesting. And so I just started building things with it every single day and uh very quickly you know the first one of the first things that I did is I reverse engineered the Mural photo frames in my house so that I could update the photos on them because the software doesn’t work. And I, you know, in 15 minutes I built a new web interface uh to update my photo frames and they, you know, it built me this beautiful interface. I updated the frames, I got new photos on them, super happy.

Jason Calacanis: 3:46 This Alex is one of the great moments that everybody has. We should name it when you get claw-pilled because this is a sub-moment. Yeah. Um and I have two moments. One is when it actually builds you a piece of software. Yeah. Or asks you shall I just build you a piece of software? Because that would be like I don’t know you’re in some incredible fine dining restaurant and they just delight you you know in the 11 Madison Park you know unreasonable hospitality way. I’ll call this like the unreasonable open claw hospitality where it said to me when I was doing some work on like a bunch of names of people in Japan that I wanted to uh do like a little uh networking with. And it was like shall I build a CRM? And I was like you go girl! Do it! And it was like pink! Here’s your CRM! And I was like I- what? Insane. Yeah. What? Insane.

Greg Kara: 4:38 Insane.

Lon Harris: 4:39 Yeah.

Greg Kara: 4:41 Insane.

Jason Calacanis: 4:42 The second moment I had and I want to know if you had this one then Alex I’ll give you a chance to to get in here because it’s just such a great guest that you and I are going to be battling for uh questions here. My- the second moment I had is when it goes recursive. And I was like okay tell me about your soul MD file, tell me about your skills. And then I had Matt Van Horn on, another old school Web 2.0 head.

Greg Kara: 5:05 Which by the way I have a Matt Van Horn story but keep going.

Jason Calacanis: 5:08 Absolutely. He also has been pilled, claw-pilled.

Greg Kara: 5:11 That might have been my fault.

Jason Calacanis: 5:13 Oh so you’re to blame, I see.

Greg Kara: 5:15 Yeah he’s all in.

Jason Calacanis: 5:16 So uh and he’s got an interesting new startup. I’ll leave it at that. He teased it here um that you know we all probably have an affinity for. I’ll leave it at that.

Dave Morin: 5:22 A little affinity.

Jason Calacanis: 5:24 Wink wink nudge nudge know what I mean?

George Yameen: 5:28 No conflict, no interest.

Jason Calacanis: 5:29 Conflict interest all- all that stuff. But I told my original agent Deckard uh ‘Hey go use this last 30 days skill’ and in the production room look at thumbnail images on YouTube and just every week on Saturday learn a skill of the best practices for thumbnail images and look at our thumbnail images and rate them and then tell us how to make them better.’

Greg Kara: 6:00 it’s even better.

Jason Calacanis: 6:01 Ah, this is a good idea.

Greg Kara: 6:02 Right, so it comes back—

Jason Calacanis: 6:03 I’m going to do this for more or less. Game on.

Greg Kara: 6:05 You should do it for more or less. Another amazing podcast, uh, with four, two great couples, four great individuals. Uh, and then it said, oh, you know, um, this VP at MrBeast did an interview on this other podcast I’d never heard about and he said on the podcast that they use heat mapping to test 50 thumbnails in Google—in YouTube searches, then they take the five best ones and then they test those using the YouTube tool, then they pick their best one in the—in the bake-off and I’m like, well, that’s a little bit too much for me, but I’m glad I know that. And then it said, ‘Should I, when I hear about…’

Jason Calacanis: 6:41 Should I just do it?

Greg Kara: 6:43 Well, yes, and build us some software and off we’ll go right to Human. But that recursiveness and then it’s like by the way, a lot of the things I learned this week I already have and it’s other people talking about the original research I found. I’m not going to include that because that doesn’t make sense to just repeat it. And I’m like, you go girl.

Jason Calacanis: 7:11 Yep.

Greg Kara: 7:12 100%. Okay, so those are the—those are the two moments for me is the recursive and the unreasonable hospitality software developer.

Dave Morin: 7:14 Well, I think the—the other thing that I really fell in love with immediately was the… there were sort of two things. Well, three. The—the memory feature, right? Like all of us using, uh, AI have used ChatGPT and the big models, right? And it has like some memory of you and you can kind of go check it out and there’s probably—there’s some other, like, voodoo magic they’re doing in the background to, like, make it so you can access your former chats but it’s like—it’s like loose and you—you never really know whether it’s going to work. And so the idea that like I’m getting this—these memory files on my own computer that I can then use and reuse and do things with and there’s like basically a journal every day, like that was number one. Number two, the skills feature, uh, which, you know, you’ve seen skills like you said Jason, there’s skills on Claude Code, like skills is like this new emerging primitive that’s happening. But the idea that you could tell the Claude bot or the now Open Claude to build itself a skill and then simply share it with the community using ClaudeHub and then other people immediately start downloading it. I did this with a snow report skill, uh, right out of the gate of course, uh, and built some other ones but this really easy ability to build your own skills and then share them with the community I think is an underrated feature. Uh, and then the third thing is this like crazy idea that, uh, Peter had which is called Heartbeat. Which is he gave the bot a heartbeat. And you know, we’ve all built—anybody who’s built software builds these cron jobs and you get the—there’s these things called demons and cron jobs and—

Greg Kara: 9:00 you can run like demons are sort of running always in the background and cron jobs run like at a certain time and do something. But the idea that like the machine has a heartbeat and it will do things each time it does this heartbeat is like really thoughtful and interesting and it adds this new layer of kind of being able to your point, like run run recursive things, like have it do things like every five minutes, every 15 minutes. And it it has this alive feel to it that I think um has uh is really really interesting.

George Yameen: 9:35 It feels like it’s looking out for you and wants to help you proactively and that’s a new kind of experience for software products.

Jason Calacanis: 12:00 Ask this person about that or remind me about stuff. And so I do production meetings now for the podcast. I pop up in my Plaud, I do the production meeting and I say here’s what I want to talk about on the pod, these are the four or five topics, and I will be talking to two producers or two executives at the firm. And then I can just summarize it and it has a whole gallery of different transcripts, summaries, action items, all different ones being created like the skills. And then I can send a link immediately to another team member and they just have it. And because it’s part of the hardware, I don’t have to open my phone, it’s so not intrusive.

Lon Harris: 12:37 Does it break down who’s talking, Jason, at any given time or is it just a blob?

Jason Calacanis: 12:41 Oh, and it remembers it. So if I say Alex, Lon, Dave for the next conversation, it has it. And this has made my life like, I’m saving hours a week with this thing.

Lon Harris: 12:51 Send me one.

Dave Morin: 13:21 Perfect.

Jason Calacanis: 13:21 Alright, Alex, you must have a question here. Sorry to monopolize the first part of our discussion.

Greg Kara: 13:25 Oh no, you’re fine. When it comes to friends, I just sit back. But Dave, the foundation was amazing news as Jason said. Everyone was very excited that there was something in place to ensure that Open Cloud would continue. So what I want to know is how early were you guys working on the foundation, who else is involved, and then of course what your goals are for the next three to six months.

Dave Morin: 13:40 You know, from the first conversation, I reached out to Peter early January out of pure gratitude, just to tell him like thank you for building this, and that started a conversation. And this was before anybody, I mean there were like hundreds of people in the Discord, and you know we now have over 100,000. And we also just passed Linux and React yesterday in GitHub stars, so we’re the number one open source project in the world.

George Yameen: 14:12 That’s crazy.

Dave Morin: 14:14 Immediately my impulse was to just help Peter like look, whatever this turns into I want to be part of helping protect it. Like it’s too important. I think Jason, like you were saying, as soon as you become Cloud-pilled, you’re immediately just like this is so important.

Jason Calacanis: 14:35 Have to protect it.

Dave Morin: 14:37 Yeah. And having been around the technology business a very long time and been through several different platforms, even having been part of building, you know, the Facebook platform back in the day, I’ve seen a lot of what can go right and what can go wrong. And so I just immediately told him like let’s get started trying to figure out if there’s a foundation here. No matter what you decide, you know, he’s been very open about his thinking. I mean, you can just go on Lex or any of the podcasts that he’s done and he was very open about his thinking and what he was thinking through. And so that was the idea was just like, let’s make sure that we have an open source project here that’s a real thing that has a real, you know, 501(c)(3) nonprofit foundation behind it. There’s a lot of precedent for this, be it, you know, Linux or, you know, any number of the open source foundations, Apache, Mozilla. There’s a lot of this stuff out there. And let’s do this right and make sure that both the technical side, Peter never loses, you know, decision making authority over and on the open source side that we set this up with good governance. Let’s try to get as many of the big labs involved as possible and try to make it the Switzerland of AI because that’s really I think the experience that people have once they start using it. They’re like, wow, I can use my own Mac as an orchestrator to interact with all the different AIs however I want. And so the foundation should play the same role in the ecosystem. And so that’s what we’re working towards.

Lon Harris: 15:59 Dave, I will get back to the major labs in a second, but tell me more about Peter maintaining essentially direction or decision making. What is his continuing role and how much authority does he wield still?

Dave Morin: 16:10 Look, it’s his project. We’re still working out all of the governance. We’re like deep in the middle of it right now. But it’s Peter’s project. And so I think you can expect that he’s going to make the decisions, especially the technical ones. And the foundation exists to protect it, not to control it, try to, you know, influence its direction. The foundation has to be in the middle of the broader community like you guys and taking input, making sure that the global conversation is heard. But at the end of the day, Peter built this thing. And so, you know, we’re just here to serve.

Lon Harris: 17:13 On the sponsorship point, you do have OpenAI, Vercel, Blacksmith and recently announced Convex as sponsors of the project. You mentioned getting other AI labs in there. Anthropic famously got a little bit twisty with the project originally over some naming rights and so forth. But do you expect to get Anthropic, xAI, Mistral and maybe even the Gemini team to all also support?

Dave Morin: 17:47 Yeah, let’s hope so. You know, we’ve got conversations going with everyone. We’ll figure it out as time plays out. And you know, we’re just excited to see where it nets out.

Lon Harris: 17:59 These open source projects, when they get this big, you get a very large community involved. What becomes the dynamics in terms of just politics?

Jason Calacanis: 18:00 Explain to the audience, because I think it’s confusing for people, who gets to or how are all the different changes—how does air traffic control work for the different changes at a—at scale project like this? Because there’s so many people wanting to contribute, obviously skills can be done independent of the—of the—the core codebase, but how does that work, and is that what the foundation needs to do, is like maybe hire air traffic control to—to manage it?

Dave Morin: 18:28 Yeah, so the way that it works is that anybody in the world can download the project. You know, you can go on GitHub, uh, you can download a copy of it, you can make changes, uh, you know, improve it, and then you could submit pull requests—or as everybody, uh, you know, you’re probably on Twitter or if you’re not familiar with the nomenclature, people call them PRs. Um, and what that means is that you’re submitting a change to the open source project and you’d like to, uh, have it included next time there’s a release. Now, when a project gets this big, the number of PRs goes through the roof. Um, I actually haven’t looked today, uh, you know, I think we’re up above 5,000 at any given time.

Dave Morin: 20:10 I mean, it’s so many more PRs than are possible to, uh, work through. There is a group of people called the maintainers. Um, the maintainers of an open source project are usually a much smaller crew, you know, I don’t know, different projects have different numbers, 20, 30, 40, 50 people, um, that watch all of the PRs coming in. They also have their own ideas for what needs to be improved, and they prioritize what’s most important. So, the most important thing for the last few weeks has been security. Uh, whenever you have a really large project that a lot of people start working with, uh, security becomes number one. Um, and so there’ve been a lot of security, uh, updates. It also, you know, in this case, we’re working on AI, and so there’s a lot of different models being released, uh, every week. I think there’s a lot of new things happening, trying to include more models, more integrations, stuff that makes sense. A lot of requests to integrate this, integrate that, like, “I want to have this part of my workflow, will you put this in the core project?” right? Like, all those kinds of things. And so the maintainers have to look at all of that and make editorial decisions as to what will improve the project. And we’ve really been in improve-the-project, you know, from a stability, security, like how it works today. We have not been working on, you know, as much the new features right now.

George Yameen: 21:41 Do the maintainers get paid by the foundation, or are they just like super volunteers that have extra, you know, merge-like privileges?

Dave Morin: 21:46 Right now they’re super volunteers. And we ideally would like to, you know, bring on some of the best, the people that are most interested, those kinds of things. And, you know, once we have the capital in the foundation and these kinds of things, then we ideally will pay them for their work to the extent we can.

George Yameen: 22:00 So Dave, how do we measure success here? Because we’ve been watching the GitHub stars go vertical for weeks now, incredibly impressive, I don’t mean to diminish that. But for the foundation, what does success look like in stewarding this project? And what’s kind of like the North Star, apart from improving the product, handling security issues, and interacting with the community? Like, where do you want it to go?

Dave Morin: 22:31 Yeah, I think that success looks like a stable and growing ecosystem, right? Like, you could perhaps look at how many projects are there out there, both, not, you know, open source that have included OpenCLaw, you could look at commercial projects that are built on top of OpenCLaw. And that we have a really just stable, beautiful, growing ecosystem, not unlike Linux. You know, I think most of the early internet was built on Linux, right? Most people don’t know that, but that’s the truth. And so, to the extent that this can be a huge part of how people build things going forward, I think that’s success for us. I also think we have to continue research, improving things going forward, making it better. OpenCLaw, at the end of the day, was a very powerful idea. You know, the idea that you can give people the power of AI on their own computer, this idea of a personal AI that actually does things, takes actions for you. I think this is a major part of the future, and we have to just continue to push towards that frontier.

George Yameen: 23:47 Is it possible to donate to the foundation yet? Is there like a donation page yet? And then we’ll get to the business side.

Dave Morin: 23:53 Not yet. Yeah, we’re hoping, you know, within the next couple of weeks, as we start to get the business side together. You probably know, anybody who’s set up all the various types of business entities in the world, 501(c)(3)s are one of the most sophisticated. So, we’re working through it as fast as we can.

Jason Calacanis: 24:10 Learning from the comment section asks, what hardware does Dave Morin recommend? What’s your stack? You in a virtual machine? You got a studio? What do you think?

Dave Morin: 24:19 Uh, I use a Mac Mini. Uh, I have several Mac Minis. Uh, and I also use virtual machines. So, you know, for my, uh, my venture capital firm, we use a DigitalOcean virtual machine that operates in our Slack. Uh, and my team there has absolutely loved it. Um, and, uh, I’ve also got a bunch of testing right now, like Alex said earlier, Vercel has been an amazing partner to us at OpenClaw. Guillermo and his team are just world-class, we love them. They’ve got a new, uh, a new virtual machine environment called Sandboxes, and so I’ve been doing a bunch of testing using Sandboxes. They’re not really set up yet for OpenClaw to be super effective. You have to keep them alive every five hours and these kinds of things. But, um, doing a bunch of testing there. But I think the answer to the question is whatever your use case is, you know.

Jason Calacanis: 25:21 I just moved mine, Dave, to a Mac Studio.

Dave Morin: 25:25 I’m hoping to do that. I was going to see if there’s new chips coming out this week.

Jason Calacanis: 25:31 There are. I don’t think they’re going to come out this week. I think they’re doing that new MacBook that’s like super entry level Chromebook.

Dave Morin: 25:37 Oh. I was hoping for an M5 Mac Studio this week.

Jason Calacanis: 25:41 I think they’re saying June for that. I talked, um, yeah, uh, which is a bummer. But I, the reason I got the Mac Studio is because I wanted to run Kimi 2.5 locally and we have that up and running. And my, uh, my replicant is using it right now, uh, so the last 48 hours I’ve been on Kimi.

Dave Morin: 25:50 How’s it going?

Jason Calacanis: 25:51 Couldn’t notice a difference.

Dave Morin: 25:53 Really?

Jason Calacanis: 25:54 I didn’t know we had switched.

Dave Morin: 25:55 Really?

Jason Calacanis: 25:56 Now that was really weird.

Dave Morin: 25:57 Really?

Jason Calacanis: 25:58 Yeah. And so here we are, folks. But I think we’re going to, um, I think we’re going to hit some…

Dave Morin: 26:00 Are you running one or two Mac Studios?

Jason Calacanis: 26:03 Just one right now. Um, but this is like a very light instance. But I find running it on a Mac, I was on a Mac Mini, I find running it on an independent hardware stack much better because when you want to write software or you want to use native things, it’s so much better. So one of the things I’m trying to get it to do is create its own Apple Photos. I’ll give you another use case. Many times I want to use a clip on the program and then I, you know, you want to download a clip and I’m like, that’s a lot of work. So I told it, hey, do me a favor. Like there’s all these like open source projects to rip, you know, an Instagram, Twitter, um, YouTube video, and they always get shut down when they’re on the web, but the open source projects are still there. So it’s like, okay, yeah, I popped that up for you. And now I send it a clip and it then puts it in my G Drive for me. But I was like, I want it in my Apple Photos so then we have a shared photo. So that’s not working yet. It’s a little bit janky. Once I get that up and running, then I was like, wait a second. I don’t know if you’re like me, but do you bookmark stuff on Instagram… like or TikTok constantly for like, when I’m next time in Japan or Paris, I got to try this. Do you do- are you part of that group?

Dave Morin: 27:18 Only Twitter. Yeah. I don’t use Instagram or TikTok.

Jason Calacanis: 27:22 Right. So… interesting. It’s too addictive.

Dave Morin: 27:26 Smart man. Smart man.

Jason Calacanis: 27:27 But anyway, this is like how I- you’re like men’s fashion, there’s like books I want to read or, you know, movies I want to see. And I’m constantly bookmarking them and then I got to remember to look at my bookmarks. Now, I send the clip to my OpenClaw over Slack and I say fashion, media, inspirational, docket for like, the docket of the podcast. It rips it. Then I gave it an instruction, Dave. Go give me intelligent feedback on whatever I send you and you determine what intelligent is. So I’m like, for like bad fashion choice…

Dave Morin: 28:05 Yeah, well…

Jason Calacanis: 28:06 and I’m like, so if it’s fashion, tell me like comparable brands, tell me about the history of this piece, because I was looking to get my sweater game going. Educate me on that. Then for like movies or books, tell me about the director, tell me about the author, tell me the Goodreads score, tell me the Rotten Tomatoes score, all that stuff. So now it’s like, okay, this is not just my bookmarks. This is my curator who’s understanding what books I want to read and what fashion I’m interested in and what places I want to stop when I’m in Japan next.

Dave Morin: 28:34 Yeah, Jason, but this is KIM-E 2.5 being the intelligence nexus that makes those decisions for you, right? So it would have the opinion on your sweater, just to be sure I’m tracking this?

Jason Calacanis: 28:48 Not exactly the opinion. What I want it to do is just intelligently reflect back to me more information on whatever I send it. That was like how I said it. And I am going to work on that skill. I started doing it this weekend, but now that I’ve got, hey, just might as well take the whole video, transcribe it, put it into a Notion page, and then I’m going to develop a skill, which is my curator. And so I’m going to create a curator skill and then boom, I’m all done.

Jason Calacanis: 30:23 On an investment level, you and I were very disappointed that Openclaw did not become a for-profit/you know, open-source thing. We’re, I’m sure we were both lobbying heavily for that outcome, but hey, to each founder their own and happy for Peter obviously. What do you think the opportunity is for investors in Openclaw? Because I created Openclaw at launch.co and said anybody with an idea and a decent team and a prototype, whatever, just email us, we want to invest. I think we’ve done two so far. I just want to invest in like 20 of these companies, incubate like 20 of them.

Dave Morin: 31:01 Yep, same.

Jason Calacanis: 31:02 What do you think? What do you think is the opportunity?

Dave Morin: 31:04 I mean, I think that is the opportunity. I think we’re at the beginning of an ecosystem that there’s going to be a lot of interesting stuff. There’ll probably, there’s already a Cambrian explosion of people doing different things. Usually, you know, anyone who’s been around technology for a long time knows that, you know, the first wave is always kind of the first ideas out the gate are simple. There’s a lot of people trying the hosting thing and the this, that, and the other thing. It’s usually the second wave where things get really, really interesting. And so I guess the opportunity is just to stay in it, encourage people to try things, try building things. I’m hearing a lot of college kids actually say like, what am I going to do? There’s going to be no jobs. And I keep telling them every entrepreneur I know could not be more excited right now about what’s going on. They have more tools and more things to build with than they’ve ever had. Openclaw is one of the best ways to build things and try out your ideas. Just start building things. And the goal of having an open-source foundation here is to enable a stable ecosystem where entrepreneurs and investors can expect that this project stays around for the long term. And so you can build things and trust that you can build things and that it’s not going to be ripped out from under you.

Jason Calacanis: 32:31 No rug pulling. That’s the key. And as we’ve seen, a lot of times large corporations will be excited about a developer program. Twitter had a very robust one for some period of time. Facebook had a very robust one. And our pal Mark Pincus is like, oh my god, this is going to be the greatest, Zynga and Facebook forever. Heart, heart, heart. And then, you know, at some point the platform, Microsoft as well, you know, oh we have WordPerfect and, you know, we’ve got WordStar. Yeah, build whatever you want on Windows. And then…

Greg Kara: 33:00 At some point, the platform says, “You know what? There’s like 20 really big opportunities on this platform. We’re taking all of them.” Now, Apple might be very generous about it, and they very slowly will say, “Oh, our notepad will add some features to it,” but we’re not going to like kill Evernote year one. They just, you know, going to keep it very simple. So great for them. But yeah, that is the challenge as an entrepreneur, isn’t it? Like, who can you trust to keep the platform open and free? And a nonprofit with a leader like you, Dave, puts the trust level right at 100 for me.

Dave Morin: 33:39 Yeah. I mean, I think that’s what I hope we’re going for here. You know, there’s a lot of aspects of the AI ecosystem where people are trying to make these decisions, whether you’re an investor or an entrepreneur, “Can I build this product? Can I build this idea? Or is it going to be eaten in a year?” Like, I hear that every week at our investment committee table and in talking to other investors. Like, people are just like terrified that the, you know, the idea that they’re working on might be consumed by the big AI models a year from now or even sooner, right? And so I think that’s what one of the positives here of OpenCola is that, you know, it’s all being built in public. You can, you can build on it, and you can expect that it’s going to be around and, you know, and, and build whatever you want.

Greg Kara: 34:27 Dave, Intuition partner Hugo Anselm, he’s said that half the current YC batch has already pivoted, and it seems that everyone’s having the same crisis that you’re talking about, like, what will work in a year? What will be alive in three years? So what’s your mental model for founders to understand what will be accelerated by OpenCola and Agentic AI and what will be utterly subsumed by it and turned into, you know, a skill.md file?

Dave Morin: 34:54 Yeah. I mean, I think that you want to stay at the orchestration layer as much as you can, and you also want to be focusing on areas where you’re creating unique data that, um, you know, that is not trained into the models, for example. Like, I’m doing a lot of robotics investing right now, and one of the… I know that’s the thing du jour, but one of the main diligence questions that we’re looking at is, is this robot creating a distinctly new data set that, um, that only it would know, um, and that therefore the company has access to either real-world data or digital data that is net new, um, in society. Um, and so I think you have to think about these kinds of questions. I think just doing like your standard, I don’t know, vertical co-pilot AI for XYZ is getting pretty dangerous. Um, and, you know, I think you want to focus right now if you’re going to do digital work, you want to focus on: what’s the agent that I’m building? What’s the… you know, am I building an agent, um, for any number of use cases? Like, can I… AI might be the best agent for this type of job, right? Like a lot what you’re seeing is people building these, you know, agents to handle different parts of their business as if they were employees. And so I think there’s going to be a lot of innovation around this. And what happens when you do that? You’re basically training an agent on something that you uniquely know, right? And so I think that’s that’s very exciting. And I guess back to the first thing I said, the orchestration layer is very interesting. We have this company in our portfolio that I’ve been working with a lot called Polsia. It’s p-o-l-s-i-a dot com. And this is a solo entrepreneur who spent the last year building an agent orchestration system that enables other people to build their own businesses. So think Shopify, but for the agentic world. You have an idea, you want to start your own small business or even maybe even a big business, Polsia will help you think through the idea, do research, do market research, it will start running your ads, it will build your website and every day you come back and it has new tasks for you to do.

Jason Calacanis: 37:21 Oh, we just had Polsia on the pod on Friday.

Dave Morin: 37:23 He’s great.

Jason Calacanis: 37:24 So you invested in this? See if you can talk to the founder about getting a very high-profile podcaster who’s good for a tweet once in a while with a big social following and a couple big podcasts to sneak into the last round for a hundy. Just a quick 100k, because I want to be—I want to do more deals with you, Dave, so we have an excuse to like just hang. So just slip me in for a hundy, or if he wants to go plus 10 percent from the last round and open a note, I’ll put in 250, maybe hundy 250. I just—I was so taken by his idea. I think it’s genius.

Dave Morin: 37:58 It’s amazing, right? And he’s working at this orchestration layer, right? Like how do you orchestrate agents? I think of it as like if agents are kind of like the orchestra, people have to build symphonies. And Ben’s built a beautiful symphony here, right?

Jason Calacanis: 38:14 I’ve been calling it the maestro. I’ve been, you know, after that movie that came out, like who’s the maestro of all of these agents? And did you solve that for yourself? Like I’m assuming you have multiple agents. I know Brit’s got agents working.

Dave Morin: 38:29 I’m still working through it. You know, there’s a lot of different ways to do it, but I think we’re all kind of thinking through the various different ways to do this right now.

Jason Calacanis: 38:35 I had one, I have two concepts. One, I have my super replicant that has the keys to the kingdom. Only I can talk to it. And it’s pretty cool when you give it access to every Gmail at the company, every Slack at the company, and every Notion edit at the company, and every Zoom. Because now when I’m doing executive coaching, Dave, I say to the replicant, ‘Hey…’ Tell me Dave’s previous week, all the emails he sent based on category, sorted by importance, every Notion edit he made, every Zoom he had, and every calendar entry. And what do you think was accomplished this past week and how should that inform next week? And be an executive coach and go research executive coaching. So I don’t know if you remember this theory, but it was Travis’s, which was everybody’s private driver, like everybody gets a chauffeur. So there’s this concept, take whatever a rich person has, private jet, give them JetSuite, their own private driver, Uber, their ski house or their house in Hawaii, Airbnb. So I was like, you know what’s interesting is executive coaching is something that’s like, what does an executive coach cost in Silicon Valley now, Dave? Like a good one?

Dave Morin: 39:47 I don’t know, that’s a good question.

Jason Calacanis: 39:49 I think 10k a month is probably the ask. Five or ten.

Dave Morin: 39:52 Probably.

Jason Calacanis: 39:53 Maybe more for the best ones. And for the best ones they’re probably at 25, and you know what, if you’re on the board of a company worth a billion dollars, you’re like, ‘Yeah, I’ll spend 300 grand a year just making my CEO 5% better.’ Yeah, totally worth it.

Dave Morin: 39:59 Totally worth it.

Jason Calacanis: 40:00 So what if everybody had an executive coach? Well, what does an executive coach do? They watch you work. So for a CEO, that’s like not creepy. It’s like Steph Curry wants every video of him shooting ever. For a rank and file employee, they’re like, ‘Oh my god, I’m being micromanaged.’ No, you’re being macro-coached. Yeah, like change your mindset. So I created this executive coach skill I’ve been refining and then I just said, ‘Give people their week in review.’

Dave Morin: 40:25 Is this an open source skill?

Jason Calacanis: 40:27 I haven’t started making these public skills yet, but I will because I would love… ‘Send it to me now.’

Dave Morin: 40:38 I want it. Send it to me now.

Jason Calacanis: 40:39 When I make it a public skill, I will send it to you.

Dave Morin: 40:41 Or send it to me privately.

Jason Calacanis: 40:42 I’ll send it to you privately and then it has a thing where it opens up your Coinbase and it just sends me a Satoshi every hour. So don’t worry about it. If you see anything about Satoshis being sent, don’t worry, it’s all good. Okay, I wanted to ask you a lightning round, couple of new stories. I know you got to go.

Dave Morin: 41:00 Yep.

Jason Calacanis: 41:01 Alex, there’s been three big new stories. I want Dave’s opinion. Davey, you have 90 seconds for each.

Dave Morin: 41:06 All right, here we go.

Jason Calacanis: 41:07 Gonna hit you, we’re gonna hit you with the hard ones. Go ahead.

Dave Morin: 41:11 Let’s go.

Jason Calacanis: 41:12 Was Anthropic correct to decline to meet the DOD’s terms as last week ended and maintain its red lines?

Dave Morin: 41:17 Spicy. Spicy. I feel like I don’t know enough about what’s going on with this.

Jason Calacanis: 41:26 Oh god, the-the-the-the-the… you can’t group chat claim ignorance. Come on, Dave.

Dave Morin: 41:30 I didn’t read much about this.

Jason Calacanis: 41:33 Okay, here’s basically what happened. Anthropic is like, ‘I have two red lines.’ Number one, we’re not ready, this software’s not ready to build murder bots, so please don’t use Claude to create bots that kill humans. And then number two, we don’t want to be involved in building a surveillance state against Americans. So please, don’t use our software to spy on Americans. If you can do those two things, we’re happy to sell to you, but we are not going to— Give you a license if you do those two things. Your thoughts on a private enterprise giving terms to the military of how they want their tools to be used and saying if we don’t need the money so you can cancel our contract. That’s what they basically said in effect.

Dave Morin: 42:15 I guess I just feel I don’t know, I tend to use an Apple frame on this, which is like if I’m selling Mac Minis to the DOD, the DOD’s going to use the Mac Minis to, you know, do what they want to do. It’s a utility, right? Like I was I did see that there’s like Starlink Mini terminals on top of these new drones that were running, right? Like are they like is Starlink selling them Starlink terminals and then saying like you can’t use it for this or that? Like I don’t know. I’m not a military expert. I don’t know like the various different uh, you know, variables going into what’s actually going on. These two points seem pretty political and kind of advantageous for marketing to me.

Jason Calacanis: 42:58 Ah, interesting take.

Dave Morin: 43:00 Yeah, you know, Anthropic has a pretty consistent past of using the doomer narrative to do marketing. And so both of these points feel identical to me, you know, it’s like oh, you can’t do surveillance of all Americans. Like I highly doubt that the DOD asked for that.

Jason Calacanis: 43:21 And they’re technically it’s illegal for them to do it because they’re not allowed to spy on Americans, the CIA, FBI we have like some, you know, division there. Okay, second story Alex tie up a second one, lightning round.

Greg Kara: 43:32 In the last couple of days we have seen OpenAI announce the new staple runtime environment and also Anthropic dropping an auto-memory feature. It’s clear that the AI labs are getting agentic with it. Do they actually challenge OpenClaw or are they too far behind to catch up?

Dave Morin: 43:49 Um, my take is that the and you know OpenClaw is your own personal AI. You run it on your own computer. That’s a very different thing than running any of the things you just said in the cloud. Um, they’re just different things. So I just I just don’t I think it’s comparing apples and oranges.

Lon Harris: 44:07 Even if you had the expanded cloud memory feature and you threw in like a computer use agent in there, to me you’re getting close-ish to the same idea.

Dave Morin: 44:17 Yeah, it’s fine. But again like if you’ve got a computer in the cloud that somebody else controls, like you don’t have control over it, you know, to the last question like that could get subpoenaed, it could get, you know, you can get all that data pulled out of it. Like I think people, maybe not everyone, but I think there’s a large amount of people that just like you like having your own MacBook and carrying it around in your backpack, you might want to have your own personal AI agent that sits on your own computer, right? And that’s…

George Yameen: 44:42 That’s the key for me.

Jason Calacanis: 44:44 This is so important. It’s so important for people to start thinking about their data, the proprietary nature of it, and do you want to and this is this is no dig to Sam, but we’ve all known Sam for a long time, he’s got sharp elbows. Yeah, but he’s got… Sharp elbows and he wants to, he has to build a trillion dollar company here. If there’s any revenue opportunity, he’s going to take it. So if you’re shipping all your data up there, don’t be surprised if you wake up one day and he creates, you know, a legal co-pilot, a coding co-pilot. That’s his right to do that.

Lon Harris: 45:15 Yeah, totally.

Jason Calacanis: 45:17 But it’s also your right to say, you know what, all my venture data, all my secrets, all my podcast secrets are on my Mac Studio and we’re never upgrading, updating them to the cloud and we use an open source project. That’s your counterbalance.

Lon Harris: 45:28 Yes.

Jason Calacanis: 45:29 Viva la difference. Last one. Alex, do you have a last?

Lon Harris: 45:31 I do. In response to Anthropic’s choice to not work with the DOD on its requested terms and OpenAI saying yes to a very similar set of terms, people have been dropping OpenAI and sending Claude to the top of the iOS App Store. Mr. Dave Morin, do you think that there’s any chance OpenAI loses the consumer AI crown to the dweebs over at Anthropic who previously have been the enterprise crew?

Dave Morin: 45:58 Alright. Well, I think it proves my first answer right, which was that marketing worked.

Jason Calacanis: 46:04 Yes. It’s like I’m giving up 200 million in a military contract and I’m going to get 10 million more users and I’ll go right to number one, which is hard to do.

Dave Morin: 46:14 It’s the Kayfabe theory of internet engagement, right? Like you have to be you have to be a face or a heel and whether you get cheered in the ring or booed, they both work.

Lon Harris: 46:23 Absolutely.

George Yameen: 46:26 Can I just say that I am shocked at how cynical we are on this point because I watched the Dario interview on CBS, I read everything about it and like I still get this vibe of they actually have some moral fiber to them.

Dave Morin: 46:46 I think they do. Like, look, I think it’s fair to give them credit like they clearly operate out of principle, clearly, right?

Greg Kara: 46:52 Disagree with it or not.

Dave Morin: 46:53 But I also think they are great at marketing.

Greg Kara: 46:54 And those two things can be true.

Jason Calacanis: 46:56 In related news, I just got Opus 4.7. I got the early release, Dave.

Dave Morin: 46:59 What? How?

Jason Calacanis: 47:01 Yeah, I got it from Dario, sent it to me directly and it says here that I have been misgendering people and that I now have a misgendering negative score and that I have to work on that and that will be on 60 minutes this week. A woke Opus 4.7—I’m joking.

Greg Kara: 47:22 Oh man.

George Yameen: 47:24 Where can I invest immediately in 4.7? Yes, get your pronouns correct.

Jason Calacanis: 47:31 Dave Morin. He’s the guy. You can listen to him every weekend on an exceptional pod with his wife, Brit. It’s called More or Less because Dave Morin is on it and the Lessons are on it, including Jessica Lessin from the Information.

Dave Morin: 47:46 Oh, thanks Jason.

Jason Calacanis: 47:48 What a great pod it is. I enjoy it.

Dave Morin: 47:51 When somebody’s out sick one week, surprise the audience and have me come on. That I think would be hilarious.

Jason Calacanis: 47:56 Alright, let’s do it.

Dave Morin: 47:57 Just throwing it out there. Just like, I’ll come on and just act like whoever’s off, just act like I’m now in a relationship with the other person.

Jason Calacanis: 48:00 Alright, let’s do it. Sounds good, perfect. All right, brother, we’ll see you soon. Great job. Great kayfabe! Great kayfabe!

Lon Harris: 48:03 We have two more guests and demos, I think, and we can keep moving here because we’re 53 minutes into a show. It’s a lot of show, folks.

Jason Calacanis: 48:12 It’s a lot of show.

Lon Harris: 48:13 Oh look, I put my plaid pen on.

Jason Calacanis: 48:14 Oh, very nice.

Lon Harris: 48:15 This is what I wear when I’m skiing, and then, you know, during the day I have my other one, and you can have two different plaids working for you. Anyway. Now Jason, I mentioned Founder University just a few minutes ago, and I want to bring up a founder that is in Founder University. This is Greg Kara of Run Tools. You might be familiar with this product. It was called something else back in the day, but now it is all about the world of agents. And Greg is going to show us how it works, spin up a bunch of OpenClaw agents, and then put them somewhere, Jason, I don’t think you’ve ever seen before. So please welcome to the show, it’s Greg!

Greg Kara: 48:49 Hey everybody, thanks for having me on the show. Hey Jason, um, hey Alex. Thank you. So, um, we’re Run Tools AI. I’m the CTO of Run Tools AI. We are a AI-native infrastructure company. It’s actually kind of crazy how much our thesis aligns with the conversation that’s been going on in this podcast. We provide workspaces, sandboxes, workflows, agents, interchangeable with OpenClaw, tools, and code execution.

Dave Morin: 49:13 Okay, so this is hosting, you’re hosting OpenClaw and making a better front end for it. Am I correct?

Greg Kara: 49:18 Absolutely, yes. Hosting OpenClaw, making a better front end for it, but also all the primitives needed to kind of complete the picture, right? Um, and if you don’t want to use OpenClaw, you can build custom agents, um, or you can use OpenClaw, it doesn’t matter. So—

Dave Morin: 49:33 So everybody’s doing hosting. How do you stand out from the thousand other people who are now like, ‘Oh, you know what I’m going to do? I’m going to host OpenClaw.’ Like, OpenClaw Foundation, maybe they’ll have a hosting option like WordPress does. So how do you think about that? Is this like a temporary business or how does this become like a sustainable long-term business?

Greg Kara: 49:54 Basically we started by building an OpenClaw before OpenClaw came out. Um, we’re software developers, so we kind of started with a coding harness and then I started adding all these integrations and I’m like, ‘Hold on a second.’ And this is around the time that 4.5 came out, where tool calling got really reliable. And I was like, ‘Hold on a second, this is really cool. Let us build something with it.’ But we pivoted— before OpenClaw came out, we pivoted, yeah.

Lon Harris: 50:09 Smart move.

Greg Kara: 50:10 Yeah.

Lon Harris: 50:13 Because yeah, it’s like you’re trying to brick by brick build something and then all of a sudden this castle gets built and there’s like a million people working on it.

Greg Kara: 50:21 Exactly. It’s like why would I want to build a car from scratch? So what do you see— this is the dashboard. We kind of gamified this intro so, you know, people that are not technical can make their agents, so you can just click continue.

Lon Harris: 50:33 Okay, so instead of having to set up your OpenClaw, this walks you through it like a wizard really fast.

Greg Kara: 50:39 Exactly, exactly. Really fast, got it done. Um, so this is the dashboard. We have our Agent Hub. This is where you create your agents, different agents, connect them together, select the sandboxes they run in, so on and so forth. Um, our Tool Hub, tools you just install tools here. Um, and we have an SDK, so everything is declarative in code, so anything you push through your dashboard—

George Yameen: 51:00 Repo also ends up on the dashboard. And then these- this is our sandboxes page. This is the sandbox we just created right here. And if you click on it, you can get access to its desktop environment real quick. So can your agents.

Jason Calacanis: 51:12 That’s nice. You have the desktop running in the web interface. Great.

George Yameen: 51:15 Yeah, you have your SSH here. If you’re a technical user, you can run SSH commands on the left and or you can watch your agent work on this desktop environment. And then if we go back to code execution, this is the endpoint I talked about. So if we click run Python, Python returns. If we click JavaScript, JavaScript returns. So all you have to do is just curl this, you know, just- just make a call to this API endpoint with the language and the code and you get the result. This is for building more secure enterprise agents that don’t need to run code on- on prem. And then these are the workspaces like I showed you. You can go into all of the folders and see the files your agent created and see what it’s doing in there. This is what I like to call workflows. It’s like a nathan type tool because nowadays these agents-

Jason Calacanis: 51:59 You kind of let them loose, right? That’s the point. You want them to do things on their own kind of. But sometimes you want deterministic outcomes. Like you want one agent to read your email, you want the other one to create a CSV out of it, you want the other one to do something else.

George Yameen: 52:10 Yeah, you can drag and drop and create flows here and connect many agents together.

Dave Morin: 52:16 Oh, that’s slick. Yeah.

Jason Calacanis: 52:19 Yeah.

George Yameen: 52:21 And then- and then this will also, once you deploy this, this will also give you an API- an API endpoint and as a skill you can install this to your open clock- so your open clock can actually use this workflow as a skill. This is our virtual office and if I start this, you’re going to see these- these agents come in and have a seat here.

Lon Harris: 52:42 So you’ve created with the Unreal Engine essentially a first-person shooter of your office. So you can run around like you’re playing Doom.

George Yameen: 52:54 Exactly.

Greg Kara: 52:56 That’s going to take a lot of CPU, but that’s super fun as a sim tool to watch. It’s like a Black Mirror episode, Alex. We get to watch people walk around and do their job.

Jason Calacanis: 53:04 Yes, but also maybe completely unnecessary. What’s the value here, Alex?

George Yameen: 53:09 Well, what I like about this is it lets you see your agents what they’re doing. And the reason why I thought this was demobable is that the screens that they have in front of them are actually what their individual screens are doing. And so you can literally go up behind your agent and look over their shoulder and see what they’re working on. And that- that to me was like, okay, cool, so this is not just a toy, it’s a tool.

Lon Harris: 53:32 And- and I’m gonna go ahead- this is exactly like the Black Mirror episode where they’re stuck in the Star Trek thing in the simulation and the guy’s like torturing them in the Black Mirror USS Callister, yeah. That was a great episode.

George Yameen: 53:44 And- and the cool thing about this is we do truly believe that this is where work is going. Each organization on our platform will get their own virtual office. So all of your agents that are doing work on the dashboard, you can come in, check ‘em out, see what they’re doing. You can go to your little CEO office here and you can control them from your CEO computer and when they have no work to do they kind of get up and kind of walk around.

Greg Kara: 54:00 round and talk to each other and share information and also if you don’t like VR and if you have like an AR headset and if you have a physical office you can bring your agents into the physical world and have them sit next to you on the desk.

Jason Calacanis: 54:11 That’s actually the big win. So this is, let me give you some advice as an expert in demoing. You start with what’s strongest. What’s strongest is this insanity and the vision of having AR glasses. So what this would mean is for our replicants we would look over in our Austin office and at the conference table during our management team meeting would be Deidre, Roy, Chris, whoever.

Greg Kara: 54:32 Exactly. Exactly.

Jason Calacanis: 54:33 And we would see them through our headset, which is nuts. Uh, this is a great vision. Well done and uh, yeah, this is an interesting vision. I give you credit for this. Good job.

Greg Kara: 54:44 Thank you very much. Thank you very much.

Jason Calacanis: 54:47 Alright, let’s drop him off. We got one more demo to go, yeah? Thank you Greg, and by the way Greg, you can come to the Launch Festival, go to launchfestival.com. We’re selling only 50 tickets. I think half of them are gone already for people who are not founders, a thousand bucks each. Then we give the other 350 tickets to our investor guests who are judging and the speakers and to founders to keep the event free. You can just buy one of the thousand dollar tickets, come to lunch with me both days, and you will support all those 350 founders coming for free. Greg gets to come for free. I got one of those precious founder tickets for you. So go to launchfestival.com. We’re doing it on March… somebody help me out.

Lon Harris: 55:23 16th through 17th.

Jason Calacanis: 55:25 16th and 17th. Monday and Tuesday, March 16th and 17th, in two weeks, in San Francisco. 400 investors, founders, and a couple of, you know, folks who support those two previous groups. Launch Fest. Alright, who’s next on the program, Alex?

Lon Harris: 55:46 I’m really excited about this one. We have George Yameen of Picklewatch. He’s building an app that helps track your pickleball swing using your Apple Watch. It’s very, very cool and he has been playing around with agentic automation and we have a demo planned for you, Jason, that’s really, really cool. We are going to show a task being done using both Open Claw on one side and Claude Chrome plugin on the other to show their different strengths and weaknesses. George, welcome to the show.

George Yameen: 56:08 Jason, what if I told you I could save you thousands on operating costs with your organization. What you’re looking at here, my screen on the left is Open Claw with Chrome Sonnet 4.5. Okay. On the right you see Claude Sonnet 4.6 with the Chrome plugin on Chrome. Got it? Have you used the Claude plugin at all?

Jason Calacanis: 56:32 All the time.

George Yameen: 56:33 Okay, so I have some…

Jason Calacanis: 56:34 Yeah, yeah. I mean, I don’t use the… I don’t use CodeX or any of the programming tools. I’ve used Cowork. I have the downloaded app. And when you have the downloaded app, a lot of times it needs to use MCP, I guess, through the Claude Chrome extension, and then it gives instructions to your browser to do things like maybe go into my LinkedIn and read my in-mail and make it into a, you know, a database of my in-mails.

George Yameen: 57:00 Whatever. I got a video demo because it’s a long task that I ran and I’m just going to speak through it speed run. On the left what you’re seeing and on both these browsers is with Pickle Watch, the Apple Watch app, I track my users live with something called Smartlook where I can literally see my user’s screens, where they’re at at the onboarding and if the onboarding is converting, yes or no. What I’m doing with this agent is watching user behavior in real time and then updating the paywall based on what’s working and what’s not in real time.

Jason Calacanis: 57:32 Ah.

George Yameen: 57:33 What this 10-minute little demo’s going to do is it’s going to go through my sales, what’s working, it’s going to go through my Google Analytics for where my users coming from, how long are they waiting, and it’s going to literally be watching user’s screens. All right, so I gave basically the same prompt to OpenClaude, you can see my Telegram window here, on the right I gave the same prompt to the Claude plugin, um, and I’m going to talk about the strengths and weaknesses. So on the right, this one jumped straight into an active session of my app of someone using it live and it’s taking screenshots, it’s clicking, you know, it’s just going through. This one on the left, I, you know, it’s just like, hey, I see these, this browser, I see these windows, whatever, you know, it’s getting to those environment. Um, and let’s click through, let’s speed run. OpenClaude is jumping straight into completed sessions, active sessions. Claude code, sorry, Claude Chrome plugin on the right is creating events custom and and tracking that way. So let’s just speed through a little bit. As you can see on the left, OpenClaude is quite chatty, which is cool. In five minutes, OpenClaude, amazing, opened up my the paywall and is creating an AB test paywall in real time that it could test and iterate.

Jason Calacanis: 58:50 Wow, that’s crazy. So, all right, just to catch the audience up, if you are a founder, you will watch user sessions. There’s different services you can do and different software you can get to watch the user use your app. Typically what happens is somebody who’s a product manager or somebody in customer support or a founder will watch those videos and they’ll look at their Google Analytics and find where like, oh, people don’t click the purchase button or the try now button. Or maybe you want to say try now or purchase and you want to AB test those two calls to action. So you set that all up manually, you do a test. What this is doing is you’ve got on one side OpenClaude looking at whatever the software you use for your testing and there’s a million different testing frameworks. And then on the right you’ve got Claude’s, which is different than OpenClaude, Claude is the product from Anthropic and they have a Chrome extension. That Chrome extension is looking at actual users live on the site then telling OpenClaude… Here’s what I learned and then open claw saying let me form a test based on that or what?

George Yameen: 1:00:06 No, they’re both it’s doing what you the latter that you just said separately and I’m just trying to show the pros and cons and Jason the open claw wins the…

Jason Calacanis: 1:00:13 Got it okay so two different ways to do it and I think open claw wins the test is what you’re saying?

George Yameen: 1:00:19 It’s faster, it’s better, it’s more customizable, it’s amazing, I’ve been Claude-pilled, mind-blowing. But Jason, here’s what I really need to hit you with. This one on the right cost, and Alex I think I corrected the math, it’s 10 to an order of magnitude of 10 cheaper than we thought, 0.3 cents to 7 cents versus on the left it cost me $16 in API calls.

Jason Calacanis: 1:00:41 So you used $16 of API calls to run the work using open claw, it worked, it was fast, it was good but it was expensive. And then the Claude Chrome plugin cost you a couple of pennies. Is that API or is that like your fraction of your subscription cost?

George Yameen: 1:00:56 It’s a fraction of my subscription and you know it refreshes every 8 hours but you know use your open claw to max out the subscription that’s a ton of money.

Lon Harris: 1:01:05 So this is just a bit in the weeds Alex but a good point and it’s great that you put some dollars on it so this is really a great demo, well done.

George Yameen: 1:01:13 Thank you.

Lon Harris: 1:01:14 Pro max plan is like 100 bucks a month from Anthropic. Anthropic probably loses money on this or some users who max it out they lose money on but people who like use it lightly like it’s just a ChatGPT they probably make a ton of money on those so that’s why they banned open claw from using your pro max subscription because that’s not what they wanted to do because they know the open claw users are going to rip through the credits and then that would turn their economics upside down. So there’s a lot of inside baseball going on here. But George what’s your story? Are you an entrepreneur? Do you run a company? You work for a big company? Which one is it? Walk me through.

George Yameen: 1:01:57 I do. No, I’m full-time into PickleWatch. I found out my wife was pregnant and I got laid off in the same week and my life coach therapist mentor at the time told me stop applying to jobs, build something. I built something.

Jason Calacanis: 1:02:12 Ah so PickleWatch is your company.

George Yameen: 1:02:13 Yes.

Jason Calacanis: 1:02:14 And PickleWatch competes with what?

George Yameen: 1:02:18 Have you played pickleball?

Jason Calacanis: 1:02:19 Yes.

George Yameen: 1:02:20 Awesome. Okay so this is going to be real easy. I have the app running here, I have a pickleball paddle and when I swing it’s going to color my watch green, yellow, red depending on how that swing quality was.

Jason Calacanis: 1:02:32 Got it okay so PickleWatch is an app to make you better at pickleball.

George Yameen: 1:02:43 Yes.

Jason Calacanis: 1:02:43 And there is one of these called Carv for skiers, Alex, which goes onto your boot and it does the same thing on a very refined so you actually have a consumer app and you just use this to figure out how to make your consumer app better. Now I get it. Okay.

Greg Kara: 1:03:00 Perfect.

George Yameen: 1:03:02 Track your swings, speed, and it gives you real-time feedback. After, this is the simulator, you could keep score in real-time as you’re playing. It’s going to give you feedback custom to your game and how you’re playing.

Greg Kara: 1:03:16 All right, so Pickle Watch, there you go. I’m looking at it right now. Well done. And so, can you make enough money here to support your family or no?

George Yameen: 1:03:27 Uh, we’re getting there. We’re having a slow grow, but the potential’s there, you know.

Dave Morin: 1:03:32 This is it? Pickle Watch track pickleball? Is this, I got the right one? Wow, it’s a good-looking app. Um, you got some good ratings, 4.8 rating. And you charge 100 bucks a year for this thing?

George Yameen: 1:03:42 Uh, no. 45 a year, $2 a week, and then $80 a year. It’s going to go up, if anyone’s curious, pretty soon so you could get it at a discount now, but it’s free to use, free to try. But that’s not, we’re going to hard paywall this pretty soon because it’s growing.

Lon Harris: 1:03:59 George, I’m just curious about how many people you’d have to hire to replace the agents you’re currently using? Because I know you’re building this by yourself and I know you also worked a lot to get ready for us today. So like, you know, if you had to hire people to replace your little, your little friends, how many would it take?

George Yameen: 1:04:14 Definitely like probably one or two marketing people. Like that paywall thing I showed you with the AB testing, I was dreading doing that guys for like three, four months and I did it in one night with one prompt and it blew my mind. I was like, oh my gosh, like so, so powerful.

Jason Calacanis: 1:04:31 Awesome. Hey listen, great job George. I think, you know, sometimes you build something you’re passionate about, which is great, obviously your Pickle Watch is great, but you may stumble onto something, so I always tell founders keep your peripheral vision open. What you may have stumbled on is that there might be a bigger opportunity for you in doing watch and app experiments, refinement and making a tool for people who build that. So you might have like an even better concept that you could pivot to because those people might pay 500 a year and you might be able to get a thousand of them and you get a thousand people you pay you $500 a year, a solo entrepreneur you make 500 grand and then you said you got a wife and kids, yeah?

George Yameen: 1:05:22 Yes.

Jason Calacanis: 1:05:23 And then they might be like, oh, okay, yeah, don’t get a job because this is going to pay you four times as much. So just a piece of advice from a, from a husband and an entrepreneur, great job. You’re invited to join us March 16th and 17th. Join us at launchfestival.com. So you, George, get to come to launchfestival.com.

George Yameen: 1:05:35 Thank you, Jason.

Lon Harris: 1:05:37 Jason, before we, before we let you go, I want to put you to the same test we gave Dave. It’s only fair to ask you the same question. So when it comes to the Anthropic DOD Department of War, sorry, situation, where do you come down on private companies and their rights versus the government and its needs?

Jason Calacanis: 1:05:55 Yeah, you know, this is America and we have a long tradition of people getting to… Be pacifists if they want, they can literally for religious reasons not join the military during a draft and not be deserters, I guess. And so, you know, if the people at Anthropic don’t want to participate in the military, and it seems like, you know, it seems like this administration has like some issues with Anthropic. So, you know, there’s a lot of gamesmanship going on here between those two parties. I liked Dave’s take that maybe this is being done by Anthropic to get press or whatever, but I just want to say, you know, it’s not anybody in America can make the choice if they want to work with the government. If I said to, with a very rare exception of like, we’re actually in a war and the country’s being invaded and then you can take over production, like if…

Lon Harris: 1:06:46 Yes.

Jason Calacanis: 1:06:47 …you’re Tesla and you’re making cars and all of a sudden the country’s being invaded and we have God forbid another Pearl Harbor or 9/11 and the government said to you, hey listen Elon, we need you to make tanks or we need you to make submarines or whatever it happens to be, you know, it’s like life or death. Okay, fine.

Lon Harris: 1:07:01 Yeah.

Jason Calacanis: 1:07:02 Makes total sense. However, you know, in the case of like a peacetime situation, let the company do what it wants to do and I don’t think they need to ban them as like some terrible contractor across the whole thing. That seemed like an unnecessary escalation. Do I think it’s patriotic for a company to make tools and then not allow the government or the military to use them personally? I think that’s a little like ungrateful. But listen, we allow people who are religious or who have like certain philosophies to hate gay people. Like I’m Catholic, we don’t allow women to be priests, or I grew up Catholic, I should say I’m Christian, not Catholic, but there are allowed to hate gay people, not allow them in the church and say it’s a sin, and they’re allowed to not let women be priests. Okay, it’s sexist, it’s homophobic, but I guess in this country you’re allowed to be sexist and homophobic if you say it’s for religion. You’re also allowed to be a peacenik and say I don’t want to participate in this. And you can also be like I just sell PCs and chips and batteries and if people want to use a battery, I don’t, I can’t tell them what to do with the battery. So both of these arguments make sense to me and I think we just need to settle this down. And if somebody’s building a War OS, like a war-based language model, what a great opportunity. Just, you know, build a language model explicitly for the military. And that actually seems like an incredible opportunity for somebody. Where is the Anthropic or OpenAI or Grok or Gemini specifically for the military? That sounds like a great thing to do. If murder bots are not ready, it would be a lot better for society if instead of sending humans into combat we sent robots into combat and they took care of bad guys. And by the way, with a bomb, what do we do, Alex? Do we send humans to diffuse a bomb anymore like the Hurt Locker or do we send a robot?

Lon Harris: 1:08:57 I believe we have many robots for this task now.

Greg Kara: 1:09:00 Precisely. Can I throw in just two cents before we wrap here? Uh, Anthropic has been pretty clear that they are super content to work with the DOD, have… are they’re working really hard to make sure this transition off of Anthropic onto, I guess, OpenAI technology is smooth. So I- I think that when we discuss patriotism, it’s important to- to not conflate it directly with just what SecDef wants, but instead what- what makes sense for the nation as a whole because I- Anthropic’s taken a lot of stick lately and I don’t think it’s entirely earned. I think there’s a lot of posturing to your point and I don’t want us as a community to say, ‘Hey, you don’t get to have moral stances simply because the DOD or DOW says so.’ So, I- I give them a little more credit than the average Twitter poster.

Jason Calacanis: 1:09:48 All right, Lon has a question. Here we go. Go ahead, Lon, you can- you can break in, Lon, and ask.

Lon Harris: 1:09:52 So Jason, would you personally fund a startup that said they were training an LLM to power domestic surveillance and autonomous killing machines?

Dave Morin: 1:09:59 I know the answer to this.

Jason Calacanis: 1:10:00 Well, domestic surveillance definitely not. Um, I don’t think that that’s a great idea. I think we have to really think um about new laws and regulations because AI makes domestic surveillance a lot different than just I happen to have a camera in my 7-Eleven in case I get robbed. Like now you can have a camera that tracks across all the Ring cameras, you know, a lost dog. Completely noble. I think that was a great product. But you could also have a Ring camera that surveils people and knows their entire pattern, and that’s wrong. Now, the cameras I have for my different uh properties, on my properties, I have cameras from Unifi um that record license plates, record facial recognition. It just comes like that out of the box, but that’s for my property. Do I want that on every lamp post everywhere? No. Why? Because it’s going to be abused. So we need to have some common sense, thoughtful legislation about that. So I wouldn’t want to be involved in that. If you explicitly said we’re going to track everybody domestically, no, I wouldn’t want to invest.

George Yameen: 1:11:05 For murder bots? Yes.

Jason Calacanis: 1:11:07 Now, for murder bots, I think there’s an- an important discussion to be had of if crazy dictator Kim Jong Un decides to send 10,000 troops from North Korea into South Korea, South Korea should have autonomous robots that could counter them and that, yes, if fired upon, would neutralize. Would I back a company doing that that was run by very thoughtful people that, you know, had a philosophy about it? Yeah, I would. Just, you know, it’s not exactly in my interest, it’s not in my uh kill zone if you will, but I do think that that’s better. And what I would like to actually see, and this is what I would love to back, is non-lethals. I’ve always thought non-lethals was a really great investment category. And non-lethals include stun guns, right? There’s- there’s nets… and webbing, there’s foams. And so I did do a deep dive on that a decade ago when there was a bunch of police killings and I said, ‘Why don’t we use these foams and stuff like that?’ Pull up, um, foams that, uh, on YouTube. There’s a foam you can spray on rioters or an emotionally disturbed person or a person with a knife or a shooter. Now, if it’s a school shooter, whatever, I don’t care, take them out. But if you have an emotionally disturbed person with a knife who’s distraught or, I don’t know, they lost all their money investing in MicroStrategy and they’re destitute, Alex. You know what I’m saying, Lon? There’s a foam that you spray on people that immobilizes. It’s an immobilization foam. This looks like something in science fiction, but they immobilize this guy with a foam. You spray the foam on them, they can’t get up, they can’t reach their gun. So, and it’s sticky, and it’s non-lethal. I think this could be like an incredible… so this falls under the category of non-lethals. What I would like to invest in, if somebody has an idea for this, I would like to do a military tech investment. I’m very, you know, pro-military startups. I’ve got some favorites, in fact, and I’m very fond of the executives who run these companies. So I would like to look for non-lethal, um, murder bots. So non-lethal bots is what I would invest in to, uh, just put it plainly. All right, this has been another amazing episode of TWIST, March 2nd, 2026. In the can, as my friend Leo Laporte would say. Sundays, watch This Week in Tech. He’s Alex, I’m Jason. X.com/alex and X.com/jason. We’ll see you on Wednesday. Bye-bye.

George Yameen: 1:13:39 Bye-bye.