In this video, the speaker demonstrates how he uses AI-driven "intelligent baby" agents to automate the repetitive task of posting Loom videos across multiple crypto community channels (CoinPicks Genesis, Army, and Inner Circle). He explains the agents' current limitations—being early‑stage and error‑prone—and outlines a workflow where he diagnoses failures, saves successful actions, and prompts the agents to analyze their own history to improve future performance. The speaker also describes a broader system involving several localized agents, including a plan for a collaborator (Jesse) to create his own agent, emphasizing that humans will shift from doing routine work to managing a fleet of agents that operate 24/7 and can be scaled (e.g., 70 agents running in parallel). He promises to document the ongoing journey and how this approach will be leveraged to “crack crypto.”
Chapters
So everybody's talking about Multbot, Claudebot, whatever the case is, and people are trying to automate their business. This is actually real. It's kind of like a race. For example, I make Loom videos. So I make like audios or videos for my community to update them on what's happening in crypto. And so there's kind of like this useless, I guess just brute force thing where I post the same video to all 3 of my communities, right? And so usually I told Jesse, I'd be like, Jesse, here's the Loom, here's the title, go ahead and post it to my 3 communities. Because it's kind of like a mundane task. So what I did, look right here, it's actually posting it for me right now. It's literally posting the link. Look, it just posted it. It might make errors, but I'm actually, as you can see right here, controlling it from my phone. So it's going— it's posting to CoinPicks Genesis, Army, and Inner Circle. So it's going through the process right now automatically, bro, automatically. And all you have to do, because it's not super smart, it's like they're like babies. They're like extremely intelligent babies, so they're not perfect. What you do is you run them through a workflow and you diagnose the issues, right? So it's like, let me try another approach or whatever. And then once it successfully does it, you save it in its memory. So if it goes through and actually successfully does this, I'm going to tell it to analyze itself and its past actions, save it exactly so that when we do this again, you won't have as many issues. That's the point of it, is self-learns. And so what we're doing now, this is kind of the next step, and I'm gonna update you guys on my journey on this because to be honest with you, like, it's gonna get crazy. We have 2 other agents over there and they're gonna help each other. So essentially we have one over there, 2 over there specifically, and we're actually doing a localized model. So it's free, like a free localized model. And then we're gonna have Jesse create his own agent. Like I told you, humans are gonna go from making, doing regular work to now they're gonna be managing a plethora of agents. So that it can, because again, they're like babies. They're like little kids. They're like trying to figure this out. But I guess the advantage of an agent is that it works 24/7, 7 days a week. Not only that, you could spin up many. So although they're a little slow, you could have 70 of them and it would be faster than a regular human. That's the point. It's unlimited. You could just multiply it. So yeah, follow for more. We're gonna update you on this journey and how we're gonna use it to crack crypto.
