How to Build AI Agents That Make Money While You Sleep

Chase Haynes is an American creative tech founder running Blok Blok Studio out of Berlin, and for the past 44 days he has been publicly documenting his journey to $10,000 a month using Claude and OpenClaw. 

In this episode we go deep on what it actually looks like to build a multi-agent system from scratch, how Chase set up his AI agent Cortana, why he structured his whole business around an agent hierarchy, and what the real difference is between Claude and OpenClaw. 

We also get into skills versus tools, cron jobs, email warmups, lead scrapers, GitHub boilerplates, voice agents, context windows, compute costs, and why Chase thinks the biggest mistake people make is building agents without a structure. If you have been curious about AI agents but did not know where to start, this is the episode for you.

**

In this episode we get into:

• What Claude actually is versus what OpenClaw actually is and why people keep confusing them
  
  
• How Chase structured his entire agency as an agent hierarchy with a COO, department heads and sub-agents all working while he sleeps
  
  
• The difference between a skill and a tool and why getting this wrong costs you time and money
  
  
• Why your context window is quietly draining your compute budget and what to do about it
  
  

Why being a little bit delusional is actually a requirement for hitting big goals

**

Timestamps
00:00 Introduction
01:33 Chase's Backstory
06:43 Claude vs OpenClaw
10:30 Meet Cortana
18:00 Skills vs Tools
23:00 Sub-Agent Team Setup
25:00 Working While You Sleep
31:25 AI Voice Agents
43:23 Where to Start
01:04:30 The $10K Roadmap

**

Find Chase:
Instagram:   / haynes2va 
LinkedIn:   / chase-haynes 
Studio: www.blokblokstudio.com/call

Find Miss AI:
Website: https://www.realmissai.com/
Instagram:   / realmissai 
LinkedIn:   / keira-nesdale-b287899b 
TikTok:   / keiranesdale 
Spotify: https://open.spotify.com/show/0QIuSAR...

Transcript

SPEAKER_01 0:00

I was in the military for about five years and I got plugged into the art scene, Savannah College Art and Design. And my degree was design and technology. Just like a fusion between the arts as well as the technical side of things.

SPEAKER_02 0:12

You have obviously set up your open pool. You've got an agent. Wait, does he have a name? Or she?

SPEAKER_01 0:17

I named her Cortana from Halo. I had Cortana. The LLM I was using for this was anthropic, and it was the it was Opus 4.6 or maybe 4.5 at that time.

SPEAKER_02 0:30

What is your advice to people that might feel like overwhelmed or that they're wanting to get started, right? They they understand that AI can make them money. They don't know how, but they want to get started.

SPEAKER_01 0:41

I would hop into Claude Chat. It's this simple one. Talk about your current systems and then talk about where you're at, um, your goals, and then at the end, basically say, how can I implement AI to automate these systems? The one thing that I hope AI never replaces is that like that person-to-person interaction. And of course that makes sense, right? Humans, you know, no one can replace a human. You know, be happier with your life and make more money, right? I think AI, it's not coming for your job, it's coming for the tasks that your job can do.

SPEAKER_02 1:20

Chase, welcome to the Miss AI podcast. It's a pleasure to have you here. And you have been absolutely blowing up your Instagram, figuring out how to make $10,000 per month with Ford and OpenClaw. You're currently on day 43, so how's it going?

SPEAKER_01 1:38

Going well. First off, thank you so much for having me on the podcast. Super excited. Um, but yeah, my journey has been a long one for sure. Um, today is going to be day 44. So after this, I'll record my next video and post that as well. Um, the journey has been amazing. Um I I saw Open Cloud pop up in the news, and I just took that initiative, hopped on it, started posting about it, caught the wave, and from that it's just, you know, it's been growing and growing and compounding. So it's been quite the journey, uh, meeting clients, meeting people, you know, people like yourself, and it's it's honestly been an amazing journey, you know, meeting people.

SPEAKER_02 2:13

So I did see that it was exactly about six weeks ago that you got your Mac Mini and did the unboxing. And so tell us a little bit more about your story behind the scenes. So you actually have a creative studio there in Berlin, right?

SPEAKER_01 2:28

Yeah, so I'm in Berlin. I've been here coming on a year now, which is crazy to think. Um, I actually run my own agency, BlockBlock Studio. So basically a marketing agency, you know, we started with websites, then we got into branding, now we do social media. So, you know, the goal is to be a one-stop shop. Um, and with that being said, though, on like the back end, you know, behind the scenes, I'm always trying to like get my systems in check, be automated, right? Not throw too much AI out there, right? But enough where like the amount of time I'm spending like significantly decreases so I can focus on like tons of other clients. So that's sort of like the background on it. But yeah.

SPEAKER_02 3:08

So before you were in Berlin, you've come from America. Like, what made the change and like were you a creative, digital person before all of this, or is this like a newfound, new learned skill, and that's where you've just decided to uh take your life now?

SPEAKER_01 3:23

Yeah, I'll give you like a I'll give you the long story in a short version. So long, long time ago, I was, you know, yeah, in the US. So I'm from like California, traveled a lot in there. Um, I was in the military for about five years, and I got plugged into the art scene, Savannah College Art and Design. You know, that's when I started learning about like Adobe, Illustrator, like the very basics, right? Photography. Then once I got out, I had always set that I would go to Parsons School of Design in New York City. So I found myself in New York City, like peak COVID, like 2020, like right when everything shut down. And that was a crazy experience on its own. Um, but I'm in New York City. I did my four years, graduated um from Parsons School of Design, and my degree was in design and technology, just like a fusion between the arts as well as the technical side of things. So it was this cool fusion, you know, not a computer science fully, right? But not a full like fine arts, this nice in-between um ground. Graduated, I got a small little job with my university, but then I thought to myself, I was like, the market right now is so bad. At that time, it was when like uh you probably remember this, like the um the coding bubble basically had popped, where like software engineers, like even before the AI as well, um, it was very hard to get like a coding job. I was like, well, time to invest in myself, put the energy in. It's gonna suck for a few years. Um, but as you know, we're hopping on the call right now, it's it's it's worth the investment. So that's kind of uh the backstory of how I got here.

SPEAKER_02 4:56

It sounds like a compounding journey that you've taken yourself. And right now, it appears that you're actually following your passion and putting yourself out there, which is one of the hardest things as a creator is putting yourself out there. Um, and but it is more and more accessible for people, you know, with a phone you can start to create. And I think that with AI and how that is sort of reducing the barrier for people to start to create content and put themselves out there if they are trying to be in that digital space. So I want to know from your like so that you've done a lot of research and study academically and also like design and arts, but also the computer science. And right now you must be seeing these two worlds collide, um, you know, the creative side with the digital worlds and uh automation as well. So, how have you found that journey? And uh what has been like your biggest like lesson that you've learned from now using AI creatively?

SPEAKER_01 5:54

Yeah, so like from using AI, like I mean, quite frankly, like the like the limits, like there are no limits. You could basically use it for anything to a degree. Um what I've learned personally, so starting out when I was in college, for example, you know, I was very much like hand coding. I worked with like Arduino, I worked with like Python, JavaScript, right? Sitting there for hours coding and coding, you know, learning how to do functions. And it was great that I know that information now. Um, but then, you know, I think it was my junior year or my senior year of college, right? Chat GPT really hit mainstream. There I am. I'm finding myself coding like this intricate, like um Arduino, you know, code base um that works, you know what I mean? Just kind of like a one-shotter. So that was like my first dive into like the technical with AI, allowing me to code and actually not just spend like a week on like a problem, right? Um and then from there though, you know, it's grown. I really focus on like my like my automations and my systems and when it comes to AI, and just um I tell it to do something, and then it um gives me back that information. For example, I have a like PDF call uh PDF creator. These PDFs I put out in my videos before I was doing them manually um in like claw code and having to um you know tweak it, right? Because I would do a one-shot, but it would be wrong. Anyways, I created a skill and then it basically allowed me to just put in the title with the day, and it had everything hard-coded in it already. So, like I, you know, simply put, I've worked on like like the coding side and then also like more like the visual with like PDFs, and that's just you know, two examples of it.

SPEAKER_02 7:37

You touched on a few things there, and I want to dive deep into that because I think a lot of the audience really are at that new beginner phase of learning around how AI can be used uh to automate and organize their workflows, their daily life, or just to you know speed up some processes that they're currently doing that are manual. Now, OpenClaw is really hot at the moment. And I mean, I only got my OpenClaw a couple more, download OpenClaw a couple months ago, and I know that you started playing with it also about a couple months ago. So are you able to sort of tell the audience, like, what is the difference between OpenClaw and Clawed Code? Because some people might be starting to, well, the way that I like to put it is that you know, clawed code was the first time that people who didn't have the really high technical skills uh would be able to be able to write and create uh programs for themselves with their native language. And that really unlocked a lot of opportunity for people, like even for myself. I am not a coder, but I'm not a developer, but I'm a process engineer. So for the first time, I was able to actually, you know, get my ideas, create a process flow diagram, create a workflow diagram, talk to AI, and it was able to create the automation process of this. And that's really unlocked like a lot of potential for people to be able to actually take ideas from their head and create something from it that actually does actual work. And now we have had open claw hit the market. So would you be able to describe that what your workflows were like before OpenClaw and what they are like now? And that I think should differentiate uh the difference between what you you use clawed code for and what you use open claw for.

SPEAKER_01 9:26

Gotcha, yeah. Okay, so I'll touch on the first one. Like, what were my systems beforehand? So, you know, with BlockBlock Studio, right? I mentioned my agency. You know, I do have a graphic designer, I do have like a uh like a web dev that I work with on the team as well. And so quite frankly, beforehand, um, you know, I'd be like, you know, his name is Kyle, but hey Kyle, like I need this. Um, I had just moved to Berlin and I needed a template of like prices, right, to present to people, right? And I was like, all right, like Kyle, like I need this, yeah, these colors, these fonts based off our branding style guideline and all of that. He's like, all right, sounds good, man. Like I'll take care of it. Um, just give me like a week, right? So beforehand, for that example, right, it took more time. And of course, you know, I'm paying him as well. He's on the team. But, you know, with the introduction of um like a skill I created, right, for building a PDF. I mean, I created that in 20 seconds, right? To build the skill, it might have taken about I would say about an hour, give or take. And then 30, you know, 20 seconds in, it's uh created, it's automatically pushed to my Telegram bot. All the images are there, I download them, I can overlay them on my Instagram. I mean, it saves me a crazy amount of time. So that's like you know, one um example of how like before and after how I sped up with that. Um, the second point though, the difference between open claw and claw code. So it can get really confusing, honestly. If anyone's watching this, like it can get really confusing real quick. Um, so don't feel lost. But this is how I view it personally. So we have claw, so we have anthropic, which is like you know, the parent company. We have claude underneath it, and then within cloud, we have regular like claud chat, right? Then we have cloud co-work, and then we have like claude code, right? So that's already kind of like breaking down your sections, and then within that, we even have different like LLMs, which is like we have Opus, we have Sonnet. Um, apparently they're coming out with like a brand new LLM, which is crazy intelligent called Mythos. I don't know if you've seen that on the news, so that's coming out as well, right? So these LLMs also are you know better at certain tasks. Like Opus is really good at coding, sonnet is good at maybe for more like descriptions, high Q is more for like prawn jobs and daily tasks. So, and then also think about it as well as like a like a box, cloud is in its own environment. There really isn't a mut like you can't really be using your own LLMs, right? For example, like Quinn and such. So you're kind of confined to their ecosystem, right? I personally love their ecosystem. I switched from ChatGPT to cloud. Um, and so now cloud is amazing. So that's kind of how I break down cloud. And so compare that to open cloud now. Open cloud is like the it's been around for a while, but like it really hit mainstream. But it was the first introduction of how we can have agents operating on our computer, like our complete OS, like our operating system. It can go into websites, files, it can go anywhere, which is very scary, scary, right? Like the guardrails, right, is another markdown file that we create for these agents, right, to restrict them from doing certain things. But OpenClaw was the first introduction of an autonomous agent that could travel across your computer, right, or your server and do anything you like any anything you want. The great thing with OpenClaw is we can use you know open source LLMs. We have Quinn, we have um Kimi K 2.5, right? You're not conf, you know, you're not stuck to using Anthropics LLMs. So if that's more of like your taste, you know, you can go for that. Of course, they also have cheaper prices, which is great. Um but I would say those are the the two big differences. And I've seen it in the news though, Claude Code is definitely trying to incorporate OpenClaw and the capabilities. Uh I quite honestly see Claude basically doing what OpenClaw does at one point. But regardless, though, you're gonna be stuck in their ecosystem. So that's how I view it.

SPEAKER_02 13:28

I agree. They're kind of already trying to do that with the Claude computer use, where it is able to use, like, you know, browse the internet and be a genetic for you and do tasks for you. But that is one of the key points that you pointed out is that you're not confined to just using anthropics models. Uh, if you use open claw, you can actually go and use open source models or get something wherever you want and pick and choose for um whatever tasks that you want your little agents to do. So I want to dive deeper on that. So you have obviously set up your open claw, you've got an agent. Wait, does he have a name or she?

SPEAKER_01 14:04

Oh, I named I named her um I named her Cortana from Halo.

SPEAKER_02 14:12

So that is great. That is a great name, great name. Anyway, so you you've set her up and um you went through the downloading process and followed, you know, made sure you have set it up securely. And now what was the first thing that you did?

SPEAKER_01 14:30

When setting it up, or in terms of just like um building, like starting to build?

SPEAKER_02 14:35

Yeah, like for like everything is set up, like you've got it all securely set, and you're like first your first conversation with it. So like when you're introducing yourself, being like, hey, Kortana, um, if that's how you pronounce it, hey for Kitana, uh, is Chase here. Uh, this is your mission, this is your operation, and or did you like already have different uh like files? Like, do you want to run through sort of like the setup file structure that you should that you should do or that you use when you first of all, you know, you have just um patched your agent?

SPEAKER_01 15:07

Yeah, exactly. Um, so after I had downloaded it, configured it, and everything, I actually took this source um and this information from I believe his last name is Finn. So Alex Finn. I don't know if you've seen him on X. Um he's he was around when it boomed. Um and basically he was giving advice to everyone saying that when it comes to it, you know, present yourself of who you are, talk about what you're doing currently, and then from there, and from there, then make it ask you questions to fill in the gap so it understands you as a person. So for example, I had Cortana, the um the LLM I was using for this was anthropic, and it was the it was Opus 4.6 or maybe 4.5 at that time. And basically I did I did exactly that. I set it up and I said, Um, my name is Chase, I am living in Berlin, I am running my own marketing agency, BlockBlock Studio. I graduated from Parsons School of Design. I mean, I gave every every bit of information that I could from a personal life. Then I gave it uh my future goals, you know, also like where I'm at currently, the future goals, um basically gave them, uh gave her, gave it, I don't know, um, a whole roadmap of where I'm at and where I want to be. From there, she asked me more questions. It, you know, then we built that markdown file up more, like the structure, right? Um, and then it allowed me then to start building, and then it would understand, you know, kind of what I was doing. That was like the first initial setup with that one.

SPEAKER_02 16:43

So when you're first chatting with it, it's kind of like a chatbot, like you're you're talking with it, it's asking questions, and it's saving this information into a file that'll refer to uh for to understand where it can go. But then how did you go about actually um, I guess this is where we can touch on like skills and uh like you mentioned earlier on, you've developed a skill that allows you to create uh PDFs uh very easily now by creating a skill and you've given that skill into your open core. So, like, how do you evolve your um your open core uh into being able to actually do a gen tech work for you when you're not even watching?

SPEAKER_01 17:21

Yeah, this is a good one because I mean it can be super confusing. Because I see like these people online, they're like, my robot just did this, or like my robot ordered me pizza. You're like, I even I, to be honest, like I don't know how to do that yet, right? And that's fine. Um but what I've learned personally is make sure, you know. So we have different markdown files, right? We have like our structure, our memory, our skills. And so when you're speaking with this agent and you're training it, right? I always use this analogy, it's like like the matrix, right, where they're coming out of like the slime, like they're full humans, but they come out of the slime or whatever, like they might not have any like memory or don't know how to do anything, but like they're there and present. So you basically have to train them. You have to tell them what to think, how to think. And so when it comes to that, um for example, a really important one that I would recommend anyone to put into their agents is like callback memory. So, you know, every time that you initiate a task with an agent, make sure that it is calling back the prior information so it knows where to start and move forward. Because if not, it basically gets lost of what step it's at, right? And those are just small examples of it. So callback memory, um what else? Um definitely interacting with other agents and sub agents, right? Making sure your correspondence is good as well, and that you are not also like um exerting too much work on an agent. And I think one very important thing as well is letting the agent know how to speak to you, right? The structure you have you do have to give it a personality, right? Like technically it isn't AI, but um how it speaks to you, in my opinion, is also very important. So those are just a few things when it comes to like building your agent as well. And then of course, we have our skills and we have our tools. Sorry about that. No, you're right. Yeah, skills and then tools. So, like skills, of course, this is you know one thing as well, like the difference. Tools is like um query query a database, um, search the web, right? They're more like what I would call like functions, right? And then a skill is what I would call like the playbook of how to approach a uh a task at hand, right? So like creating a PDF, right? That's kind of like the skill is because I've created that that playbook, right? Of like branding styles, like the colors, the fonts. Um, but then the tool would be like create PDF as a function, a little more simple.

SPEAKER_02 19:50

So wait, can you dive deeper on that a little bit? So, for example, you've just set up your open claw, you've hatched it, you've given it information around you creating a um creative studio and where you wanted that creative studio to go. And perhaps the first task is to build a PDF. So did you write the skill first, or did you write the tool first, or did you write the skill and told like the skill is talking about like this is the structure of what a PDF looks like, this is the branding of the PDF looks like. Like, how do those two complement each other in the example for the PDF?

SPEAKER_01 20:23

Gotcha, yeah. So when I built it out, um basically I built it out, built it out as a complete skill. The thing, the great thing with these LLMs, of course, is um sometimes you don't have to like specifically create the tool and the skill. You know what I mean? Like it's kind of sad, you know. If like if um I was trying to give an analogy, but I can't think of one. But basically, it's like if you create the skill, they will know that certain tools will be needed to complete that skill. So for me, I built a skill, I said, um, built this PDF builder. Here's my logo, here's my fonts, here are my colors, um this is the total amount of uh pages I would like per PDF, right? And then from that though, they understand how to like create a PDF and like run like the tool, like the background functions, which then result in the production of it. So sometimes it's like automatic where it's just just plugs in.

SPEAKER_02 21:20

Cool. And I want to dive a little bit deeper on another thing that you mentioned there was subagents. So you have an agent and subagents. Can you describe uh the setup for that, how you hatch subagents, when you should hatch a subagent, and how do they collaborate together?

SPEAKER_01 21:38

Yeah, this is a really good one. And it can get like at this point, you can just build as many agents as you want. Um, but it's very good to be intentional with your agents. And so a good mindset that I have when creating these agents is just think about it as like your actual like team. Imagine if you had people, you went to work, and you had people you interacted with. That's how you should view this as well. And I know it's simple when you say it out loud, but I feel like a lot of times people are just creating these agents, you know, willy-nilly. Um, so when I created mine, I have, of course, me, so I structured mine as like a business structure. So me as the CEO, underneath it, I had Cortana as the COO, right? She's like leading the charge on it. And then from there, so as a marketing agency, right? Each each business is different, right? Whoever's listening to this, your business is different, different than mine. So you need to structure it differently. Um, but for me, I had four department heads, right? So me, Cortana, and then four department heads. I believe it was like head of marketing, tech chief. Um, I think I had a support ticket. Link in on the last one, my apologies. But basically, though, I had four, like, they were like like the um like head departments, they were like the top dogs on it. From that though, so for example, marketing, um, we have cold outreach, we have people who are writing the emails, um, we have you know, potentially people who were like calling, right? So anything that falls into marketing, you then create your sub agent that deals with that specific task. Because if you are telling Cortana to do all of this, the amount of work that she has to do, you know, it's just like a real human being. Like they're gonna get super overworked and they're like, this isn't even my job. Like, I don't even know how to do this, right? And you're expecting them, right? And that's not so you have to set it up properly so each employee that you hire knows how to do that task, right? So your agents are like the the top one, like Cortana, and then from there you build your subagents, which then can of course work together. And if you train them properly, they know how to like work together properly.

SPEAKER_02 23:48

I've seen sometimes when people have like their whole entire agentic environment, like you say, where there's a CEO, many employees that are different agents with different tasks, and then every hour or two hours or whenever you program it to be, they come together and they collaborate. Have you got the same sort of setup where um I've also seen it where people have like a whole like a mission control page where you can actually um track all of the different subagents and like give them metrics so you actually know that they're doing productive work and they're not just sitting there? Like, how have you set it up to for yourself as the human and the surveilling the whole entire land? Uh, how have you set it up to make sure that these agents are doing their actual tasks? And um, is there like a visual representation of that that you've like figured out?

SPEAKER_01 24:36

Yeah, yeah. So I created my own mission control. Um I use codex for this as well as like the base layer. But initially I set it up. Um, simply put, when you know, if we're looking at the visual, I don't have the picture in front of me. It's imagine just like a tree branch, like you're breaking down, right, with the lines and everything. Um on that though, I did um add a feature where there would be a little green button or a yellow button. I think I have red, yellow, green, and basically each color indicates if it's working or not. So red, I believe it's like look, I think it was actually just yellow and green. So yellow was an active, inactive green was like currently working on a like a task at hand. So I didn't have any like super detailed metrics, you know, like no like speed tests or like anything like that, but I had a visual um indicator that allowed me to see if that agent was working. And it's I would say it's really important to make sure you do that as well because that also means token cost. So like make sure you're also just like not letting your agents work without you knowing, and then you're like, why do I owe $500, you know, to this LLM? So I added that for that.

SPEAKER_02 25:47

Um say are they working 24-7, or is it you you sort of give them a task, say, hey guys, you need to figure out how to run a campaign that is going to generate me uh 10 leads for my um creative studio here in Berlin. That might be the task that you'll give Quintana and does she delegate it out, or how does that sort of workflow work?

SPEAKER_01 26:11

Gotcha, yeah. Okay, that was the second part. So um for that, I um I had two cron jobs. So one cron job was my social media scraper. So um, and you're probably very familiar with cron jobs, basically it's just like um like a daily task that you tell it to do something. And so I initiated two cron jobs. The first one was a social media scraper. So every day at I believe it was 6:30 or 7 a.m., I would get a text on my telegram because I hooked up my telegram Boswell, which is like a complete game changer, definitely needed. Um, and I would wake up to the top three big news of that day and that morning. So, of course, I'm living in Berlin, the time zones are different from the US, but I would get you know very accurate um, you know, things that happen. Like, for example, like Anthropic and Open AI went to India for like their like AI summit, um, you know, tons of events that are happening, right? Like Mark Zuckerberg with like the meta glasses in court. Like there's a bunch of stuff going on in the world that you need to be aware about. And so this scraper allowed me to have that. And of course, then I would use this for content um as well. So it was good to inform myself and then to create content. So that's one example. Another cron job that I find very important as well is um when you assign it to Cortana, it allows your agents to basically study the internet, research, and then find solutions that you need for your agency. So if they know you're a marketing agency, you know, you set the time whenever, right? You know, say 12 o'clock at night, then at that point they're browsing the internet, they're researching, they understand you're a marketing agency, and they're like, How can we improve this agency? And you wake up, and for me, I set it up where they would basically have the research um in like um like a project section, and then I would read over the summary of it, and then I would either like it or you know, dislike it and approve it based off if I needed it or not. So, you know, basically talking about what you were saying, like working while I'm asleep, those were two things that I had running for my systems.

SPEAKER_02 28:17

And how do you take these sort of systems and are you using them more so to you to grow your own agency, or are your customers actually wanting these kind of setups for themselves?

SPEAKER_01 28:31

I think this is the coolest thing. So it really is in my favor since I'm like like in my marketing agency, because um, for example, if I work with a client who wants, you know, I do social media now, right? And say someone's like, I I want to do social media, I don't know how to do like the hooks, right? But I understand my niche and my content, right? Say it's tech or say it's fashion, whatever, right? Because I've created these tools for myself, I can now push these tools onto like my clients that are coming in. So not only has it helped me, but now I can just not stress too much about helping my clients when it comes to getting that data as well. So, you know, I say to people listening to this every tool you can create, think about how you could scale it and then use it with clients or like people you're working with. Um, but yeah, to answer a question, that's that's how I go about it.

SPEAKER_02 29:20

It's such a hack because you're learning, uh activating and building yourself, then trying testing and refining it on yourself. And then when it's ready to go, it's ready to push. So, like, for example, you've got your PDF creator, um Boogity Boo. And if anyone's listening here, check out Chase's page. He's got so many PDFs packed with information. Uh, they are yeah, really good. I've read a few of them. So thank you for sharing all of that. So, how would you then transfer this skill, like this like um skill of being able to create the eight um the PDFs? You can do it for you, and it's like hyper-personalized for you and in your social media and your content. How would you transfer this into say you get a client who is a um a cafe, right? And they like come to you and say, Hey Chase, we want to now create like a weekly newsletter um talking about events in the community and uh what's the I don't know, baking of the day or whatever they want to be talking about. Like, how would you transfer that sort of uh knowledge that you've created?

SPEAKER_01 30:24

Yeah, so this is really good. This is basically I would so I'm using GitHub. Um, you're very familiar with this as well. Um and I'm building repositories. So I'm creating boilerplate repositories. So I actually have a really good one for like website building, right, for clients. And then of course, like the PDF one. So if a client came to me and said, hey, like I need to create um, you know, slideshows or like the bakery, you know, kind of like visuals, um, and I need to do it three times a week, right? I would then go into my GitHub repository, I would fork um like my boilerplate, and then from in there, I would then fill in all of the information from this company, right? So this boilerplate is the structure, but there's no like what I say, like meat and potatoes, right? They don't know the colors, they don't they don't know anything about it, they just know how to set it up, right? So then I plug in the information from this bakery, right? I send images, right? Just to help really get that branding style set up. And then it creates that output, and then it's its own separate repository with that. And then, of course, if I'm working with this client, I'm probably also doing it for them, right? So they're not too technically plugged in. Um, but that's how I go about creating content for other people, right? So I create a template or boilerplate and then fork it, and then I push the person's information into it. And I have one very similar for my websites. I'm currently building um a website for a client right now here in Berlin. He's a coach. Same thing. I have my boilerplate, the structures, right? It's like a like a one-pager website, and then I fill in the information, his logo, and then it builds that out as well. So it saves me a lot of time with that.

SPEAKER_02 32:06

Well, that's exactly the point of uh being able to create AI. You create the structure and then you fill it in uh and hyper-customize it to your clients. And I think that a lot of people look at AI and think that it's magic. Um, unfortunately, it's not magic as much as we wish it was magic. Uh, it does just uh speed up some of the processes that um were lethargic before the AI, right? You still need to know what you're doing and still need to work through that process. One thing that you did mention as well is that uh voice uh using agentic voices. And I want to know, have you uh given your AI agent a voice herself? Or are you uh is she able to make phone calls, or how have you integrated voice uh into like automations?

SPEAKER_01 32:53

Yeah, I haven't set up a voice for Cortana, uh Cortana specifically, um, but I did, you know, I had someone reach out to me on Instagram and they were located out of South Africa. They were trying to like sell Vindy machines, but they wanted like a a voice agent to call these businesses to like pitch their sales for them basically. And so took I created a demo for them. Um I used Twilio, I used um like you purchase a number on there, and then on top of that, um 11 Labs is really good because you can also clone your own voices. But I did create my own version of that, right? Not linked with Cortana, but I did set up basically my own AI, um, my own AI voice agent. Um, and then you know, I was on there and I, you know, um commanded it, I guess, to call me, picked up on my phone, listened to the whole pitch on my phone right there. If it was off a little bit, I would tweak it again at the code, it would call me again, fix it a little bit, you know. So um it's really cool what we can do with like the voice agents.

SPEAKER_02 33:54

So how good did you think that it was? Or like could you tell that it was an agent? I mean, you're probably trained to be able to hear like the metallic y noise uh when you're talking with a voice agent. But do you think to the standard person, would they know that it was an AI or was it they're getting I know they're getting pretty good. Uh, even I had like the other day I created a video which is 100% AI generated, and again, I just clone my voice on the 11 labs, right? And um somebody sent me a message saying, like, that sounds like oh, like you you got your cousin to do such a good voiceover. Like, that's not my cousin. I know I have a Kiwi accident, so does she, but that's not my cousin, that's AI, mate.

SPEAKER_01 34:34

That's crazy, yeah, yeah, yeah. Um, my personal experience with it. So, of course, 11 Labs, like their models are like very well trained. Like, they're like one of like the leading, the leading people on it right now. Um, for this demo, I used a really great resource called DeepGram. And this is actually really cool. And for anyone building, they actually give you this is not sponsored at all, but they give you $200 actually to actually use their system to build um you know to start building these agents. So I use Deepgram just for the demo, and if they liked it, then of course I would have like scaled up to 11 labs. Um, so my point with that is when you're choosing your service for the voice itself, make sure you're like picking one that has been trained better, right? So I would lean towards 11 labs. Deepgram is great, but it's a little more you can tell for sure. And on top of that, outside of the voice, I one thing I have to say about this is that the delay that is also a thing that you have to take into consideration. You hop on a call, there's like that awkward delay where it's and you're like, okay, this isn't human, you know? If you're able to close that gap um outside of the voice, just uh how you know you and I are speaking, right? It's flowing, it's conversational. I'm not sitting here pausing five seconds and responding, right? Um you have to make sure that your agent is also speaking with that rhythm. And of course, the voice helps with that as well, but those are like two aspects that I really focus on when I build out and I train my voice agents.

SPEAKER_02 36:04

So, how do you get that rhythm?

SPEAKER_01 36:07

So, for my troubleshooting, I've gotten that rhythm just through um like setting certain delays. So I'll hop on, you know, I'll do the training call, and then there's like a 10-second delay. I'm like, cut that to five seconds, and they cut it to five. I'm like, okay, that's you know, a little too slow still, cut it to three seconds. So that immediately I pick up, it says, Hey, so that's a lot better, right? As we train it to respond at a proper time. And then, for example, if you say, you know, you've probably come across as two, you're speaking to this phone and you say something, um, and it repeats, it's like, it's like, sorry, um, I can help you. Is there anything else? You know, is there anything else I can help you with, right? Like saying these prompts that aren't proper, you also have to like make sure your agent knows how to respond and think as well.

SPEAKER_02 36:57

Yeah, you do need to train it up a little bit. So it's like uh any any human as well, you know, if you're bringing on a new employee, you need to train them up to know what to say and get familiar with the questions in the context of the conversation that it's gonna be having. It's not gonna be fantastic on day one, but you'd hope by day 30, after 30 days of training, uh, in the real life context, uh it's gonna be pretty good at making those conversations. And the same with AI, you can train your agent, but probably in not 30 days, but perhaps every three days uh if you focus on that. So uh yeah, I think voice agents are really exciting, um, what they can do for you. Uh, and also when you start to hook them up autonomously with lead generations, I think that's when it gets really interesting. Um, and it surprises me that um a lot of people can't tell the difference between AI generated voice and AI generated content. Um just the other night I was sitting uh watching TV with my stepdad, right? And uh it was a voiceless, faceless, I was sorry, not voiceless, it was a faceless uh 15-minute video about um I think it was the um JFK assassination. He's fascinated with that stuff anyway. Uh anyways, I was going through and uh I knew straight away because it you I could pick up on the voice. And uh I sat there with him, you know it got about seven minutes in. I said, Did you know this is all AI generated? He said, What? Like, how do you know? Like I can hear, like hear like the tonality of the voice, and he's like, ah. But like it really pointed out to me something that I think is really important is that if the content or if the entertainment or the experience is good enough or providing value to the person who's watching it, I don't think it really matters if it's AI or not. Um, that's the argument that I have when some people were like, ugh, AI generated yuck. I'm like, well, sometimes it's actually better than human-generated stuff, you know? Or is it able to like give them value? If they're still getting value from it, does it really matter if it's generated by uh a computer or a sequence or by a human? I I don't know. How do you feel on that?

SPEAKER_01 39:04

So I guess, yeah, it depends on like what market, like for example, like like a commercial or like an ad or like TV, if it, you know, you might know it is, if someone doesn't know it is, but it's still providing that value or that information, you know. I honestly agree, like it can definitely be like a viable thing, like it's totally fine. Um for like when it comes to calls though, what I've learned also is, you know, when you hop on the call and you listen to them, a lot of times when the agent actually admits that it's a voice AI agent, the people are not as like upset or annoyed about it. Like there, of course, there will be people that are annoyed, right? But everyone kind of like can pick it up to a degree. And so when they hop on the call and they hear the agent and says, Hey, you know, I am BlockBlock Studios personal AI assistant. You know, we just wanted to follow up about um a meeting, you know, next week. They're like, okay, cool. Like, I'm not, you know, it's not like you're trying to be a person, right? Um, and so like that's one aspect I've thought about as well, right? Admitting that the AI agent is present in the conversation, you know, allows the person to like reciprocate better with it. That's just my personal experience with it as well. Also, a lot of legal rules for that. Um, that's a whole another conversation. Um, different continents, different countries, they have their rules with it. So, for example, in California, um, you're required to admit that it's a personal AI agent within like the first 30 seconds. So that's also something to think about as well.

SPEAKER_02 40:35

Yeah, I think telling people that it is an AI agent, it is yeah, giving them the opportunity to actually understand this and not feel like they're getting tricked. Um and when they're getting tricked, then they might be less receptive to the service. Um, if it's like, hey, this is the AI agent, and then like, okay, cool, or else they might be like, hang up and goodbye. Um, but again, I think if they're providing that value, I think it's great. So that's voice uh sort of what ways that AI can, you know, businesses can use voice to generate leads and whatnot. And um another project or pro sorry, not project, another product that I have been using or playing with before a lot of core code was NAN. And so NAN for the listeners is a software that's been around for a while now. And it was able to create process flows and uh execute these flows if it's like a repeatable task. So, like if you think about manual processes, like say writing an invoice uh for somebody, it was just like going through and executing this whole entire task and workflow. Now, with how fast AI is developing and with the rate that new tools and models are being upgraded, it almost feels like, in my perspective, that Flawed Code has kind of made NAN a little bit redundant because you can even now ask Ford Code to write you NAN workflows and it does it, which is again unlocking things. So I want to know have you played with NAN and where do you think that now that sort of automated workflow systems like fit into the whole entire AgenTix space?

SPEAKER_01 42:12

Yeah, so this is a good one. So um I've it's a curse, but it's a blessing. I've always been the type of person to like build stuff up from like bootstrap, right? Like if I want it, like I don't want to pay the subscription unless it's a very complex one. All right, that's fine. Um, but when it comes to like N-A-N, like as you know, like it's just built upon like Python, like Python um scripts to a degree, like plugging in with like other um like other services or MCPs. Um so when it comes to building them with with cloud, for example, I mean I'm building out like the full-on like Python, you know, the Python scripts, and I'm plugging it in with like, you know, I'm connecting it like fully, fully on the back end. NAN is great because it's like very much like a visual representation of what's going on, right? We have like different nodes, we have our you know, our connectors with that, right? Um I would say for like non-technical people, it helps so much, right? If you have like this trigger, right, then this trigger causes this to happen, and then this information sent here, and then you know, it could be a contact form, you know, trigger submission, and then the end output would be you have the person's name, number, and email in your um Google Excel or Google Sheets, right? So I definitely think it's there, but like I said, with Cloud Code and even like building an open cloud, wherever you build, you know, a lot of these services are kind of becoming obsolete to a degree. You can kind of just like prompt and be like, hey, build me an automation that does this. You know, you don't need to pay, you know, what is it? N A N is like $20 a month, give or take, right? Um so the capabilities to create your own automations are like ever increasing. You just kind of have to know what your goal is, and if you can articulate that well, you can create that to. Python script, you know?

SPEAKER_02 44:01

So which brings me to the question like there's so much saturation of these different AI tools, or which models or companies should we be using? One of them's good at coding, one of them's good at image image generation, one of them's good at creating workflows. Like, what is your advice to people that might feel like overwhelmed or that they're wanting to get started, right? They they understand that AI can make the money. They don't know how, but they want to get started. And then they come and they start looking, and it's really confusing. And there's OpenClaw, there's NIN, there's Claude, there's Claude Co., there's like GBTs, there's just uh, you know, then you get into all of like the image generation side of things. There's so many different models like VO3 and Cling and Nano Banana, like it's just so much. And um, and I I feel it even. I'm working in it every single day. Uh it's like a constant AI fatigue that you experience in this industry, right? And and I am in it and you're in it, and I'm sure you experience it. I mean, this is why you've built an AI agent to give you the updated news every single day because it's just changing like that. Um, so what would be your advice to somebody who like wants to get started? They perhaps want to start like an AI agency or a social media agency or build out you know, chatbots for businesses. Like, how would how do they how do they get started?

SPEAKER_01 45:24

Gotcha, yeah. First, I say to anyone who's listening to this, take a deep breath, slow down. Like you are always going to be behind to a degree of what's going on, right? Like, if you're not in like like the board members' meetings at like the top dogs, you know, like like the offices, like like you're for sure gonna be behind, right? So just relax with that, right? That's just my you know personal opinion on that. Um, but when it comes to starting though and understanding, right? You know, I think for this it'd be good to maybe come from like if you're non-technical and you don't really know where to start exactly. I would honestly recommend diving into claw, right? As amazing as open claw is for like that versatility of choosing your agents. A lot of times you need, you know, when you're coming into a new scene for AI, you need to be within kind of like these confined walls, right? It helps you understand like the basics of it so you're not getting lost in the sauce. And so I would hop into Claude, I would hop into Cloud Chat, right? It's the it's this simple one. Like uh cloud co-work is a little more advanced, right? Because it deals with your operating system and your computer, and then cloud code is like you know, like uh repositories, right? Which for that you do need a lot of technical skill. So I would start in cloud chat and I would just go into and be like, hey, you know, this is who I am. I would create a project first off as well. Projects are really good because it holds the memory in there. Start a project, say, hey, you know, my name is Chase, I run a marketing agency. These are my current systems for um, you know, shooting out emails or doing invoices, right? Talk about your current systems and then talk about where you're at, what you know, um your goals, and then at the end, basically say, How can I implement AI to automate these systems? And then from that, the LLM, right? These LLM, like this LLM, like Opus 4.6, if you're using it, super well trained, right? Like you don't have to train this agent to think, right? It's already packaged, it's ready for you. So if you're new to this, it's perfect, right? You don't need to worry about if the information is good or bad. So once it talks to you, it gives you that information, then you can start to, you know, at that point, it's you know a rabbit hole, and then you dive deep into each section, right? Um, but that's like a good start point. And that's what I would recommend to people.

SPEAKER_02 47:44

You mentioned about memory, and I want to ask yourself a little bit more on this. So, because AI agents, again, they're not magic, unfortunately. And um, they do kind of have a little bit of Alzheimer's, right? So they only remember a certain amount. So, how much memory uh do AI agents hold? And have you found any ways to optimize your memory?

SPEAKER_01 48:08

When it comes to actually working with the memory for the agents, I haven't reached a part where how do I say this? I think I've like, I think dealing with memory for many agents is kind of always going to be a thing. Of course, a lot of them, like you know, a lot of great companies have tuned it down very much to the T, but like hallucinations with these agents are always going to be a thing, regardless, right? I mean, these agents are like, you know, black boxes, sometimes the output, we just don't know, right? So the best that you can do for your memory, though, is just setting up a proper structure of like, you know, caching, understand how you know your agents work with other agents, callbacks, right? Understanding like future goals. Um, so I haven't like reached like a memory limit. I think that's what you said. But um I think dealing with the memory.markdown file, that's how I would go about it with that.

SPEAKER_02 49:04

And how about for your your context window? So this is like the when people are talking with the agent and it remembers stuff in that sort of conversation of that context, right? And the more that you chat with it, the the conversation gets bigger and bigger and fades out. So the the relationship between the most recent conversation is a lot stronger compared to the relationship with the conversation that it had, you know, earlier this morning compared to this afternoon. So it's like the relationship is weakened as that sort of window expands. And also because when your context window is getting so big, it's actually costing more compute power uh to go through all of this. So, like my recommendation to people is like, you know, a new idea, new context window. Like keep it, keep it concise, keep it short and sharp. And if you do want to transfer, say, some of that conversation into a new uh context window, ask the agent to like, hey, create a summary of what we've been talking about, uh, get the main points, bullet point, format it, you know, we don't want to miss any anything important, and copy that paste and say, hey, in into a new agent or new conversation and remind it of where you're at in a very you know concise format, and then you can start again and it's those stronger relationships. And uh I think that's something that I I've been using. Are you good doing something similar?

SPEAKER_01 50:29

Um when I'm building, I usually do build on cloud code and then like implement an open cloud, but it's very much the same um situation. Like I'll be on cloud code, like typing things up, and then at a certain point, you know, it like starts to hallucinate a little bit, or quite frankly, it says like you know, maxed out like new session needed, and then you'll just create a new session. So, like I said, with new people as well, it's really good to dive into that because it also helps guide you right to like those um those steps you need to take. But yeah, 100%. No matter where you're going to be building or coding or prompting, um, that whole context, you know, from like the memory, um 100% a thing.

SPEAKER_02 51:10

Something you mentioned there, I want to dive deeper. So you mentioned that you're actually building with clawed code, and then when that tool, um, perhaps it's a social media scraper, perhaps it's the lead generator or the PDF creator, you've built like a standalone tool that's you know is able to function and work. And then do you package that up and pass it across into your open core to then disseminate to her her different agents, whether it's a marketing agent or like how does that process flow go for yourself?

SPEAKER_01 51:41

Yeah, 100%. So this is a good one. Um so on OpenCloud, for example, I mean, it's like the wild, wild west, right? When you build, everything is based off of your knowledge, right? You're using the LLMs, right? But you have to like start parsing things together. And so with that being said, my strategy for being efficient with my time as well is I build on cloud code. So I create my repositories. For example, um, I was very excited about this project. Um, as you know, we have like smartlead.ai or Apollo.io, right, for like finding the leads, then smartlead is for like the campaigns and such. And I was so tired of like paying $90 a month. So I created my own version of it, for example, right? And I also implemented my own agent on it. So my goal was I created the whole UI um, you know, on Claud Code, plugged in like an agent as well, right? So all the systems are running, everything's great. And then once I had fully audited and like tested it, I would then push this onto my Mac Mini, right? And then it would be running 24-7, right? And then um telegram notifications and such would come to me because it's you know running 24-7 compared to my laptop that I'm on. So I build on Cloud Code, the repository, take that repository, and then I push it onto my Mac Mini. And that's then when like that 24-7 service kicks in.

SPEAKER_02 53:03

24-7 service does sound a little bit expensive, and I always get questions around this one on my Instagram. Like, how do you optimize for costs?

SPEAKER_01 53:12

Yeah, no, 100%. So, like when I say 24/7, um, it basically boils down to like um, I think I mentioned this earlier, cron jobs, for example, right? Cron jobs are basically, you know, telling it when to do certain tasks. And so I'll I like I'll put the um, for example, it was called uh easy reach, right? This is like the email outreach software that I created. And so I'll put it on OpenClaw. And so I'll set the cron job to maybe check every like two hours for like you know, incoming traffic. And then, of course, though, every morning at a set time, like currently I'm doing email warm-ups. I don't know if you're familiar, like to not be put into the spam uh box for your emails, you have to make sure that your domains are properly warmed up. So I have cron jobs that every morning initiates a um like emails sent to directly to each other's email so it warms them up, right?

SPEAKER_02 54:08

Wow, wait, wait, tell that a little bit more. What do you mean? So you've got lots of wait, tell us more. Sorry, email. Yeah, I don't know about this one.

SPEAKER_01 54:15

Totally. So this is like a really important one. This goes for like this is like literally any business, this is like business tip, right? Plumbing, marketing agency, you know, whatever. Um, when you and you get them all the time, like these emails, right? And they're like, hey, this is me, and it's like a little soft nudge email to get more interested. When it lands in your inbox, you have to make sure that the email is warmed up. If not, it'll go to spam, right? There's a lot more. That's like the simple gist of it. So for my system, though, I have cron jobs that basically initiate these email warmups. You send the emails between each other. So there's like a history of emails being sent, right? And it shows Google that these emails are like valid, that they're just not like spammy emails that were created and then sent out immediately. So I have Yeah, yeah. So I have like a cron job that initiates that. And then after about three to four weeks, then everything's warmed up and you're good. And then you can start doing actual outbound cold email outreach. And then that's when I would put it into my open claw, and then I would have sequences set up based off of like my email list, and I also have a lead scraper on it as well that would automatically scrape using a cron job as well.

SPEAKER_02 55:28

So that is a great hack. And I didn't realize this. I love the context, like warming up the emails. It's like have give it have a little gameplay to begin with. Um and then you're you're getting, yeah, I love that. That's a really great piece of alpha. Wow. Continuing on these little secrets that you're sharing with us, or what is your biggest hack uh that you have found, or biggest tip for somebody setting up OpenClaw uh today, whether it's save them memory, save them cost, uh, this skill is a definite mouse download. Like, what are some of your hot tips for anybody who is uh looking to yeah, let's keep it on on open claw, just open claw, open claw hot tips.

SPEAKER_01 56:09

Open claw hot tip. Hmm. Give me one second to think about this one because there's so many out there, and it's I definitely have a few.

SPEAKER_02 56:19

Trust us, we we're we love learning. We want a couple.

SPEAKER_01 56:22

Um when it comes to open claw, the biggest tips, um biggest tips, and we spoke on this as well, but I think it's very uh important to reiterate is make sure that you're structuring your business properly, right? Uh don't create agents that are just causing bloat into your agency. Think about this as in real life, right? If you need um someone to write emails for your business, you're not gonna hire someone who doesn't know how to do that, right? Um, they might have something cool about them, you're like, oh, this person might not know this, but does this. You're like, but it doesn't actually help with uh moving your business in the direction right where it wants to be. So that's my first tip. Make sure you're structuring your um your hierarchy um with your agents and subagents. Another really important tip also is this, okay, very important with cost, right? This is like a big thing as well. There are so many open resources out there when it comes to LLMs. So Quinn, Kimik 2.5, um I was working with a few other ones. I can't think of the names, but um when we think about it, right? Because we think about like anthropic, like quite honestly, their LLMs are like one of the best, in my personal opinion. Big anthropic guy over here. You can use those LLMs 100%, but the amount, like I don't know what the actual cost is, right? But per but per API call, right? It's say 20 cents, right? You do 100. I mean, you know, like it can really start to stack up quick, and then you're when I set up OpenClaw, I think I spent like two, three hundred dollars, like the first like two, three days, you know, just like learning burning through, I did the same, man.

SPEAKER_02 58:01

It's like I was like, it's not working anymore. Why? Oh, need to top it off again, and then then you learn. I think you already have to you need to go through this teething stage though.

SPEAKER_01 58:09

Exactly. So learning the budget. So great. If you have money 100%, I would use Opus, you know, use those tools for sure. Um, but if you're trying to go on a budget friendly, you also have to break down different parts of your open claw, right? You have the brain, you have your coding section, you have your visual section, you have like an AI section. So there were like there's like five categories on your open cloud that you should think about, right? You have your overall, like um, like your main agent running everything, but you should have five different um five different LLMs that target certain things, right? One LLM is really good for website research, like Deep Seek, for example, right? But Deep Seek wouldn't be a good LLM that could do coding compared to like Opus or Kimi K 2.5, right? And you have to review and research those benchmarks, right, to see which one fits your needs. So that's like the budget side. So all the money in the world, you know, stick with anthropic, it's very simple, right? And they're they're very powerful, but budget friendly. Look at um like China has like a bunch of really cool um like open source LLMs, very affordable as well. So that is like a very important tip as well. So budget.

SPEAKER_02 59:28

Um and then making sure oh, sorry, just carry on on that one for anybody that is looking for alternative models. Like, where do you find them?

SPEAKER_01 59:39

There's no like one set website, unfortunately. Maybe that's the next idea for for me. But um I would 100% go on X and start following people that um that are just talking about it, right? Like, you know, like Peter, for example, like Stein Steinberger, I believe. Yeah, Steinberger. Um follow him, go through his followers, see who he's following, right? Like start to connect with people and see what people are posting. On Instagram, for example, follow like the AI accounts, right? They're always posting news. And then of course you can create your own um your news scraper, right? Every morning at 6 30. You wake up to a telegram message with um, and you can target it. You can be like, hey, every morning, give me all the updates for the newest LLMs that have been released, and you'll see them, right? And that's one way you can stay up to date with it. Um, there's no like one set website, like I said, it's a good idea, but that's how I would stay up to date, and that's how I do it personally when it comes to finding what's going on for these LLMs.

SPEAKER_02 1:00:40

I I agree, that is a good idea, and perhaps you can build it for us. That'd be lovely. Thank you. Very nice. L L E M uh supermarket, right? Or something like that. Um, but so yeah, okay. I want to also ask you like there is a lot of um fear around OpenClaw, and I think because it there were people who were downloading skills that had um malware in them. So um that was yeah, not very good. And do you recommend like people download skills? How do they check a skill before they download it onto their computer, or just scratch that and build your own?

SPEAKER_01 1:01:20

So, yeah, I think this goes into like just like personal and like where you're at. Um, for me, for example, I mentioned earlier that I'm very much like a bootstrap builder, right? I see something really cool, I want to build out on my own, and I build it to the best of like my ability with it, if not like I mimic it. So that's like usually my first move into a set thing when I want it. Um there's a really good repo. Um, so it's a remotion. I don't know if you're familiar. This is something that would actually be really nice for you, but it's basically basically motion design that's um auto edits based off your video input, and it can also create videos based off text as well. So um that's one skill I created, and um you know you can tune it to fit your needs specifically. But you know, I put in my video, then remotion with the GitHub, plugged in with my skill, right? And created custom to me that creates videos. So my point to that is know uh where you're downloading it from, make sure it's verified and it's like credible, right? Like 100% use Reddit as well with that um before you download anything. Of course, if it's on like Anthropics like skill site, right, you know for sure it's been verified and it's good to go. So, like verified sites, you don't have to worry about it. But yeah, open source stuff, understand who's pushing it out. Um, you know, don't um someone message you about a GitHub repo on Instagram, probably don't download it, you know what I mean? So just be careful and be vigilant of you know where you are on the internet, these repos.

SPEAKER_02 1:02:56

Yeah, you gotta remain street smart. And uh, I mean, even I get it, you know, people send you random links. Don't don't just don't don't do that, you know. Don't click a random leg, it's probably stab and they're probably gonna hack you. Just don't put yourself in that situation, you know? Exactly. Um, but uh yeah, I think you've got to be gotta be clever about it. Now, we've spoken a lot around like how you've designed the agents and how you're building out all of these different multi-agent systems, you're giving them skills to complete individual tasks. And I think that the overall arch that well, what actually caught my eye when I found you on online there was you've set yourself this mission of uh $10,000 a month uh using OpenClaw and Claude. So I want to know from yourself, like how have you roadmapped this journey? And is it is the main source of income looking to come from clients that you're building automation systems from? Or do you set can you set up some sort of like passive income around this? Or like how do you actually achieve $10,000 a month? That's a massive goal for a lot of our listeners here.

SPEAKER_01 1:04:00

Yeah, so I I say to everybody, be delusional, right? You cannot be successful if you are not delusional. So it's it's a mindset, right? Before the technical, it's a mindset thing that you have to plug into. And unfortunately, I am a little bit delusional. And so, right, when you put a number out there, right, you know, it pushes you to achieve that. And so, yeah, today would be day 44. But yeah, when it comes to like working with these clients, um, I mean, every day I'm building, right? I'm building new tools, I'm building like client portals, whatever the thing is. And when I'm building these tools, I also it allows me to pitch them to clients that are onboarding. For example, I spoke about I'm building a website for a coach right now, right? I have a client portal that I do, it's all automated, the onboarding process and everything, right? It brings them in, and I have my own portal for his information, his logins, etc. Right. I told him about this and he said, like, that would be so nice to have for my clients that I train. So my point with that is when you are building your things up, if you can understand that that like that system that you've created can you know free up more time for them or just like clear up their pain point, right? And provide them value. You pitch it, you know, very subtly, be chill about it, right? It might hit. And then from that point on, you know, at that point you form a relationship, you're building more, you can charge more, and then of course, you know, the numbers go up. Um, but yeah, as I've been posting content and getting more out there, I've even had uh two brothers actually reach out to me and they want me to do a month-long course with them um as well, and actually, you know, teach them about open claw and cloud code. They want me to teach the basics and then build them a Pokemon web scraper. Um, and that would be a month long course, right? And so, of course, you know, that's paid as well. But like I said, it's not about the money, but these opportunities start coming in once you start building um, right? And I built a web scraper for just finding client leads, right? That same tool can. Be easily twisted than to fit just targeting Pokemon, right? Prices and margins and then cross-analyzing uh uh prices and stuff like that. So all the tools you build can be used for other people. Just know when to say it and when to put it into action.

SPEAKER_02 1:06:18

I love that you're building a library of your own skills, own tools, own agents, actually. And then you can repurpose them. And I think when you actually are talking with clients, they and they bring to you new ideas and you learn from that. You make your own agent better to like for them. Um, you know, and they bring, you know, this time it's Pokemon, next time it might be the building industry, next time it might be again the cafe coming back. You know, like all of these different industries have different purposes. And when I start to speak with people around like this is what AI can do, or this is how I use AI, then they go, hmm, do you think it can do XYZ? And I'm like, yeah, probably. I haven't I haven't tried it yet, but like I'm sure it can. When you stop to think about it, you're like, yeah, no, I can do that. Like I I haven't done it before, but I think I'm I can figure it out, you know, and I think that's the beauty of AI. Like, if you start to understand the basic, the fundamental building blocks of how to build out an agent, you know, how you should structure it, how you um can like, you know, optimize it as well. And then you can, you said, like, have a boilerplate room, and and then, you know, there's there's all these skeletons that can just be built upon depending on the different application. I think that's one of the biggest beauties of of AI. And it it touches people globally and it touches every single industry. And I think that at the moment it's sort of at a bit of a pivotal moment where uh there's so much AI out there, people are gonna get a little bit afraid. They know that it's coming, they know that they could use it, should use it, and it will probably be useful, but it's like where to use it. And sometimes there's that massive pushback, like, we don't want AI because it's gonna replace us. And I say to those people, sorry, it's here, and it's not gonna replace you. It's not gonna replace humans at the end of the day. Uh, but it can like it's gonna displace a lot of jobs, yes, but it's also gonna create so many jobs. Uh, and it's gonna give uh power back to the people to to be able to actually do jobs that they want to do rather than doing the boring, repetitive manual tasks, which give you no satisfaction, give you no dopamine. You know, like why should people be doing that when they can actually zoom out from the really basic, you know, A plus B equals C level, and then they've got the whole entire alphabet to play with, you know? That's where I think that AI is coming in, and it's like, yes, it's changing, it's new, and it's people don't like change, but it's like, hey guys, it's here, like check it out, you know, don't be so afraid of it. So um I want to know from you like, how do people go? How do you create these tools uh and content as well? So it's not just AI slot, because that's like another pushback that I get, you know, you're building all of this stuff and uh it's just AI slot. Do you have like a hack around that?

SPEAKER_01 1:09:09

Yeah, when it comes to AI slob, like I mean, quite you know, honestly, when I think of AI slob, I think of like videos and stuff like that. But I think like bigger picture, when you're building these AI tools, you know, be intentional with what you're building, right? Your time is very important. So make sure you're building tools that are actually like pushing your goals to where they want to be, right? Because it's so it's so easy and it's so fun to like create a little tool that I don't know, tracks something, right? But it's not going to actually help you in the long run. Um, right. So pick something that actually pushes you in your direction in terms of like that AI, AI slop and stuff like that. That's like outside of the videos. AI video slot is a beast of its own. But um yeah, that's how I would view it though, with that. But you know, it's inevitable. It's going to come here. We see the data centers being built across the globe, right? It's it's gonna happen, right? Like there's no way it can stop. Just like the internet, right? People were like, oh, the internet's so scary. This is before my time, quite frankly. But from what I've heard, you know, internet is so scary. What is this? Right? Look at us, we're on the internet, we're you know what I mean, connecting across the globe, right? And so AI is like that next thing. And so I think there's a mindset shift to this, right? Do you view it as a negative thing, like, oh, it's going to take my job, like, oh no, woe was me? Or do you seize it as an opportunity to become better and to stay ahead of people and to you know, be happier with your life and make more money, right? I think AI, it's not coming for your job, it's coming for the tasks that your job can do. So unfortunately, there's a lot of jobs that the tasks in that job can be replaced, like um a receptionist to a degree, right? You know, emailing and making calls. That job can easily be replaced by AI now, right? So it's not the job, it's the task. But unfortunately, a lot of people are, you know, let go because of that. But unfortunately, it's the progression of just how life goes. So all I say to that is start learning AI now, apply it to yourself and how you can get ahead of the curve so you're not left behind.

SPEAKER_02 1:11:20

I agree what as well as moving it from that task-based stuff to a process, you know, a whole workflow, like what can you actually do? Look dynamically, don't just look linearly, look like around the whole entire um, yeah, in every dimension that's possible. So And in saying that, I want to, and I know we we're taking uh a little bit of time on this one, but I want to ask one question, which I I like to ask everybody that comes on on the show. And uh this is the closing question. So, like given that AI out there is um is everywhere and people can kind of do whatever you want with it. I even saw the other day, this is actually a really cool thing, and I I want to try it this weekend. Um it was uh creating music by using gestures of your hands, and I was like, whoa, I was on like one of the anthropic ads or Claude ads, right? And uh I saw this little reel. I was like, whoa, is that actually possible? So they they put it out, and it's like that how big your hands are depends on like the sort, like how big the bass is or whatever the tune is. So um, and there was a guy literally like doing this with music, so you can get like really creative with it. It doesn't just have to be, you know, to automate a an invoice or to um you know create a web scraper, you can actually like be like apply your own hobbies into like creating with AI, like have a think, have a play, and see what you can do.

SPEAKER_01 1:12:43

Exactly.

SPEAKER_02 1:12:44

I'll let you know how I get on with that one.

SPEAKER_01 1:12:46

Please do, keep me in the loop.

SPEAKER_02 1:12:48

I will, but so anyway, like the question goes as um what is one thing that you hope uh AI never replaces?

SPEAKER_01 1:12:58

That's a very good question. The one thing that I hope AI never replaces is that like that person-to-person interaction. And of course that makes sense, right? Humans, you know, no one can replace a human. But I do uh I do hope that when it comes to these interactions, right, between AI and people, we still understand who we're speaking with, right? You know, this is a really good question. I've had too much time to think about it, but I think that would be one good thing. Like if I was to go into a work center and work with someone, I would don't think I would want them to mimic a human. I think I would be completely comfortable with them being in AI like automation, and I understand that and I'm speaking with them though, like keeping it like very transparent, as I said before with like the phone calls, for example, right? Hey, this is Chase Haynes Blockbox Studios AI agent, right? When we establish that connection and we understand like you're a part of the team, but you're your own, you're your own thing, I think that can help create like better cohesion, right, within your team. So yes, I hope that AI doesn't completely replace humanity, I guess is like one way to put it. I don't think it will, but mimicking, right? Making sure that we understand, you know, we're working with basically at one point a species in its own. Um and that's a whole different topic of like transhumanism, right? And then implementing AI into like cyborgs and robots, right? That's that's the next step as well. So that's another progression you have to think about as well, right? How real will they be?

SPEAKER_02 1:14:35

I know it's frightening what I see. Uh some of the developments of the humanoids, these guys are doing backflips, doing jujitsu, uh crazy things. And I I love robotics. It it it'll be one of my dreams come true to go to a robotics convention in China, like, oh my god, that would just I think my brain would melt. You know, I would be so excited about everything. My mind would be like going crazy. Um, but yeah, perhaps I can do it one day.

SPEAKER_01 1:15:04

Yeah, no, no, 100%.

SPEAKER_02 1:15:06

Well, anyway, thank you so much for coming on. Where what is your handle on Instagram andor BlockBlock Studios, or how do people get in contact with you if they want to uh find out more?

SPEAKER_01 1:15:15

Yeah, yeah, thank you. Yeah, so my handle is Haynes2B A, H A Y N E S, the number two V A. That's my personal. And then, of course, I have my BlockBlock Studio, B-L-O-K, B-L-O-K Studio, all one word. Um, those are my two channels. Um, come check us out, you know, come see our work, come watch, you know, videos. I'm posting every day. I haven't missed one day of posting for almost six months now. So very consistent with posting and pushing out information and making sure everyone is up to date with what's going on, giving tidbits of information as well. And, you know, Kira mentioned the um the PDFs. Every day I put out a PDF for this 100 day series I'm doing. So come check it out.

SPEAKER_02 1:15:56

Yeah, check it out. And thank you so much again. It's been a pleasure for the past hour to chat with you, learn from you, and I'm sure that our audience will definitely pick up some tips. So thank you again.

SPEAKER_01 1:16:08

Yeah, thank you so much, Kira. Thank you for having me on the podcast.