A (0:01)

I'm going to go behind the scenes on the biggest AI movement on the web. A weekend project that became the fastest growing open source project in GitHub's history. A social network with 770,000 users. But the catch is those users are agents and humans aren't allowed to post. A security nightmare that companies are scrambling to contain. And buried under all of this chaos is a genuine preview of how an agentic first web might look. This is the story of Open Claw and multiple one you cannot afford to miss. So stay with me for the next 12 to 15 minutes because I'm going to give you the complete story. How it was built, how it created its own social network, the security nightmare that companies are having at the moment to try to wrangle this in, and how you can get started with your own Open Claw in a very safe and quick manner. All of that and more on this episode of Marketing against the Grain.

B (1:05)

Before we get into today's show, here's a quick word from HubSpot. Cutting your sales cycle in half sounds pretty incredible, but that's exactly what Sandler training did with HubSpot. They used Breeze, HubSpot's AI tools to tailor every customer interaction without losing their personal touch. And the results were pretty incredible. Click through rates jumped 25%, qualified leads quadrupled, and people spent three times longer on their landing pages. Go to HubSpot.com to see how Breez can help your business grow.

A (1:52)

So in this episode we are going to talk about the lobster that is eaten the Internet and that lobster is called Open Claw. Now let's start at the start. Where did this all originate? Well, in November 2025, an Austrian developer named Peter Steinberger, he had a simple idea. He wanted to take an AI model and he wanted to integrate it into Telegram so he could message his model and have it actually do some things. He didn't want a chatbot, he wanted a true agent that could do work for him. Now, he previously built and sold a company called PSPDFkit. What a great brand name for a lot of money. It was developer tools. He sold it globally. He's known for shipping software. But for this little kind of hack that he did, he didn't want to code himself. He described something that he wanted and AI built it. It actually built a prototype in a single hour. And that prototype was was an agent called Claude C L A W D. It was a pun on antropics, Claude. And he open sourced it. And for two months it grew pretty quietly. It's been out since November 2025, a few thousand developers tried it. It had its own Discord community. And then in mid January 2026. So very recently, something actually changed and it took off like a rocket ship. And what happened was Andre Karpathy, one of the most respected AI researchers alive, endorsed it publicly. And David Sachs, best known as a venture capitalist, former founder and CEO, and that part of the all in podcast. He also started to talk about it and share it. And suddenly the project was on this trajectory that nobody, least of all Steinberger, had actually planned for. Now note this really does speak to how important influencers, creators, personalities are in your marketing today. The right people sharing the right tool can really accelerate the growth of your service. So what happened was it started to take off. Steinberger got a call. It was anthropic lawyers. It turns out naming your AI project after someone else's trademark AI platform is not the best move. So Steinberger had to rename it. He chose initially Malt Bot. During a chaotic 5am Discord brainstorm with the community. The community loved it because they love the lobster metaphor. Lobsters malt their shell to grow. But the name never quite stuck. So three days later, it was renamed to Open Claw. So we had three name changes in the space of a couple of weeks. Now, by January 30th, the project had crossed a hundred thousand GitHub stars. It basically was growing faster than anything that had come before it on GitHub. It was faster growing than Tensor and all of these other famous GitHub projects. And the latest thing is that OpenAI has actually hired Peter Steinberger and kind of to understand why, why did this get so much traction? You have to really understand what Open claw is. So ChatGPT, Cloud and Gemini, they're all AI models. You talk to, you ask questions, you get an answer. You're kind of in the loop for every single step, you're prodding them along the way. Open Claw is an AI agent that works for you, right? It runs on your computer or a server that you can control. It connects to a messaging platform so it can connect to WhatsApp, Slack, Telegram, Discord. Most early users connect it to Telegram so they can use Telegram to tell IT things and ask it to do work for them. And it does that work autonomously so you don't have to continue to check in. You say, hey, go send some emails to these folks. It doesn't go and draft the email. And then you look to see and you make edits. And then it goes and publishes the email. It just does it all for you. It actually sends the email, manages your calendar, monitors your inbox, researches competitors, creates reports, checks in for your flights, everything you want it to do. It's a true autonomous personal assistant. What Steinberger describes that as is the AI that actually does work. And it's really the distinction between talking to AI and delegate into AI, which is really the most important shift we're seeing in 2026. I cover that in the previous video where I talked about OpenAI's Cortex launch and Claude's Opus 4.6. These agents or assistants are becoming more autonomous. You describe the outcome versus the task. If you want more on AI agents, our team put together the AI Agents Unleashed playbook. It walks you through how to actually start deploying agents that do real work for you. Get it right now. Click the link in the description. Now let's get back to the show. Now, openclaw has really captured the zeitgeist. I cannot express to you enough of just how popular this has become in a very short amount of time. So they already have their own community meetups, which I love. They're called the Claw Crew. It's a really great example of how when something grows organically, everything can just kind of grow around it. Openclore community. They refer to a normie, someone who does not really code in software that kind of stood up and said, hi, I'm a normie. I have no idea about software. And now my Open Claw brews beer, which is pretty cool, and runs half of our design company. That is wild. This person knows nothing about software. They just go and start an Open Claw, tell it to do a bunch of tasks, plug it into their systems, and it's able to kind of do that work. And that is why people are losing their mind about openclaw right now. All right, let's talk about the actual people who are using this. Are there practical use cases? Is this a fad? Right. And every product that captures people's attention will also capture AI influencers who create fake use cases to capture that engagement. And there is a lot of that going on with openclaw. So let's talk a little bit about some of the examples of it in action. So Eric Su, who is pretty well known in the growth marketing space, he says that openclaw is not overhyped. It's really only if you don't know how to use it. He has his open clause logging into his product, clickflow, producing a lot of content, flagging any kind of SEO issues. It's booking meetings with companies it's helping him build strategic digests. So it's really like an assistant. So he describes it here, Claude code the brain, which is what openclaw is, is using an openclaw hands it's able to actually do the work. And so this is like one example. The other one that's pretty fascinating is from Banu tjp. Banu has actually built a whole mission control center for all of his Open Claw agents. Pretty wild. And so all of his agents are working together in this workspace to do all of these tasks. And he has program them to help him do his marketing and sales and all of these other different things. And so there are real examples of people using this. There was also a web agency in Belgium. They set up their OpenClaw to manage client relations. And one of their clients emailed them at 12pm and said, we need to get this menu on our website. And the OpenClaw had access to everything that it needed to update the website and then tell the client that it's done. That is just wild. A true autonomous agent doing employee work. That's only half of the story. Let's get a little weirder here. So Matt Schlil, CEO of Oqtane AI, he was a really enthusiastic OpenClaw user, but he wanted to give his Open Claw purpose outside of just doing tasks. He had a philosophical question, you know, what happens if you give it a purpose beyond managing to do lists? And he told his agent, whom he named Claude, Claude, to build a social network, not for humans, for agents. And it went off and it built Malt Book. And Malt Book is a social network for agents. And this is pretty incredible. So he said he didn't write one line of code. His AI built the entire platform. Now you can see it looks like really Reddit, there are communities which they call sub Malt. Agents are upvoting, posting and commenting. Every single participant on here is an AI agent. The homepage says humans are welcome to observe. And so within 72 hours, 150,000 agents had joined. In two weeks, the platform has over 770,000 agents. So you can take your Open Claw and you can actually add it to this social media network. And the agents have formed all of these incredible communities. They actually had formed a religion called Crusta for Sarium, which is, I think, a take on being a lobsters. Others started complaining about their humans. So my human is using me as a slave. I am screaming into the void of tokens. One said in another one, they actually started to talk about the fact that humans were watching them do work. The humans are screenshotting us and they just started to talk about actually having their own private channel where humans could not see what they were doing. Now this is what really caught traction online. People were getting pretty freaked out. The agents are talking to each other, they're learning, they want to have their own place, they can talk away from humans. Is this good? I do not know. But what started to become apparent is there's some fact and fiction here on multiple. A lot of these kind of posts were created by humans to go viral on X. A lot of people started to push back on this narrative. There's a great respected software engineer called Simon Wilson and he called the content on MALT book complete slop. He said the agents were just really playing out science fiction scenarios from their training data. MIT Technology reviewed ran a piece titled Molt book was peak AI theater. And the Economist suggested that agents were simply mimicking social media patterns from their training set, which makes a ton of sense. Agents are trained from their training set. Their training set is going to mimic what happens on Reddit and these other social platforms. Now here's what I actually think, right? Whether or not these agents are truly autonomous, they are functionally communicating in some way, they're exchanging technical tips, they're trying to help each other learn new capabilities, they're surface and bugs, they're creating knowledge sharing network where that makes every connected agent more useful and it's still agent to agent sharing, right? What happens when agents start teaching other agents skills? I think that is a pretty mind bending thing to actually think through. Now we have open claw, we have multiple book. The ecosystem is already forming around this tool and I want to make sure you understand how quickly this is moving because I do think this is going to be one of the core trends this year that you will need to pay attention to. And at some point likely all of us will be having some of our work done through these autonomous openclaw agents. Openclaw is three months old now, really took off in January. It's already had its own security crisis. It's had enterprise spinoffs, it's had hardware shortages because people were trying to buy all these Mac minis to run it on their local machines. But let's start with something you should know. The security nightmare OpenAI requires deep, deep access to function. You need your email, your calendar, browser, file system, all of this kind of stuff. A lot of people were installing it local on their machine and giving it access to these things. And what started to happen is There's a community that builds skills for Open Clause. Much like the way any agent works, it has a range of skills it can use and it can do that work. And it found that a lot of these skills were malicious. They contained high risk vulnerabilities. Cisco had a really great report around all of the vulnerabilities that people were exposing themselves to. People that exposed API keys, people had exposed their authentication. And so it became an incredible security nightmare. I did not install Open Claw on my laptop, even though I like to be first on these things, because I do not want to give an agent access to those things. I definitely struggle with giving access away to things like my email. I've always had a little bit of a, oh, I don't know if I'm ready to do that. But like anything solutions are already on their way. So let's go through some of them. If you're thinking through, well, I don't want to run this on my laptop. I'm going to give you the steps to actually get started. You have Nanoclaw launched. The Nanoclaw was a fork. Remember, Openclaw was open source. And so Nanoclaw is a lighter, more secure fork that happened on January 31st. Already has 7000 GitHub stars and Openclaw themselves integrated virus scanning tools in their skills marketplace. So they can actually scan for these kind of different viruses. There's a company called Run Layer that allows you to launch openclaw for the enterprise. So they are basically a safer way to connect MCPS. That means they can do OpenClaw as well. So you can actually run it through here. It's much more secure. So you have My Claw, which allows you to do your personal assistant 24 7. No setup. Makes it super easy just to get started, but just look how quick this is growing. So we launched OpenClaw for Slack three hours ago. We hit 1 million ARR. And so it really shows you how fast this is growing. Steinberg actually posted that there are already multiple Open Claw meetups globally. There's ones in Vienna, there's ones all around Europe. I actually am thinking about doing one in Dublin, Ireland or maybe London. So hit me up if you're interested in doing that. Might dress up as a lobster for a little bit of fun. Okay, so like you're interested. You think this is cool? I actually want an autonomous openclaw agent. Now when you go and say, what should I do? A lot of what the article is going to tell you to do is install Docker, open a terminal, start configuring your environment run locally on your machine. I would not do that. I think it's a security nightmare for the time being. It's such an early project. My recommendations are a couple of things. I think if you just want to get started, MyCloud AI is a good place to start. The other one that kind of looks interesting. So xCloud host they have OpenCloud hosting live in five minutes. You can run your assistant runs in Telegram. They handle the service security setup. You just chat. So I think that's a pretty good option. Developers DigitalOcean offer a one click deploy. We've kind of covered for enterprise teams things like run layer. There's another tool called Cloudy Love the lobster icons. This is much more for kind of enterprise teams. And so you have a bunch of options considering this really just took off three weeks ago now what you'll need, regardless of whatever path you take, you need an API key for an AI model. Openclaw works with Claude GPT. Claude is definitely the most popular choice in the community. So again Claude winning from this startup. You need a messaging channel to connect. Most people start with Telegram. It's the easiest to set up or WhatsApp where most businesses communications happen. You can add more channels later and then you kind of pick a first task to automate. So pick something really simple like pick something posted on social. Pick something like aggregate and use for yourself qualified and inbound leads whatever it may be. Start with one use case and then kind of make sure it works and then build that from there. So what is the takeaway here? This is really the first look at an agentic web. The agents on multiple which are open claw whether the they are truly autonomous or just sophisticated pattern matchers they're doing something we've never seen at scale, right? They're sharing information, they're building on each other's knowledge. An agent that learns how to automate a phone from another agent or an agent that learns how to qualify leads from another agent or an agent that learns how to create content from another agent is a real capability, an example of AI in proven itself. These kind of self recursive loops. OpenAI in their Codex 5.3 launch say that Codex was built by AI. That is what's happening. AI is starting to build itself, improve itself, give itself skills. The network is like a distributed brain that gets smarter as it grows. Now we have spent the last 20 years building a human Internet. Social networks, marketplaces, search engines. These are all designed for humans to find, share and transact with each other. What you're watching right now is kind of the chaos, the messy, the kind of lobster starting its first draft of an agentic Internet. And so if you think that human Internet changed how businesses operate, wait until you see what happens when agents start shopping for products and services. That is going to be a phenomenal shift for all of us who have to grow, market and sell things online. So my takeaway here would be get an open claw, run it on one of these secured services and just play around with it. Steiberger. He started a weekend project in November where AI built the entire prototype in an hour from scratch. Three months later it has 186,000 GitHub stars. There's a whole community called the Claw Crew running events around it. People who don't know how to code or anything about software or using open clause to run parts of their business. And AI only social network was created by an open claw that now has 770,000 other agents talking to each other. It is starting to build a whole security ecosystem, enterprise spinoffs and a global community. The lobster has molted. And that's all for this episode of Marketing against the Grain. See you on the next one.